Hacker News new | past | comments | ask | show | jobs | submit login
Issue 914451: Autofill does not respect autocomplete="off" (chromium.org)
820 points by bullman 7 months ago | hide | past | web | favorite | 367 comments

A few years ago, I left a $1000 tip at the restaurant up the street because Chrome filled out the tip field with my zip code (which thankfully merely defaulted to max $1000 instead). The tip field was off-screen, and the ordering software didn't have a confirmation screen, just a "we just charged your card $X amount" screen, which made my eyes boggle.

EDIT: Looking at the original March 17th, 2015 bug, it would have been at exactly around that time...In fact, checking my emails, this happened on March 18th, 2015. I had ordered from them several times before this with no problems (they used "chownow.com" for their ordering backend).

This has a lot of really serious implications. I built a form for a charity that allowed users to buy a subscription but include an additional donation amount. Chrome was sometimes filling that field with the two-digit year. The charity got a lot of complaints and it ruined the trust relationship with the donors who didn't understand what was happening and thought it was intentional.

Chrome has other behaviour that I think violates a sort of trust relationship. One of which is that Youtube would ask you "do you want to install Chrome"? Almost as if your current browser is not "what you need to access Youtube". This is especially a problem for elderly people who often use the web but don't really understand how things fit together (the way 5 year olds actually do).

> Almost as if your current browser is not "what you need to access Youtube"

This is not just confusing. It's intentionally misleading and unethical.

And it doesn't just affect old people, or YouTube wouldn't have come up with it. Lots of young people grew up with computers and understand how to do what they want to do, but they never develop a systematic understanding of what they're using.

It gets worse, though. While we tech-folk know that all modern browsers are supposed to have near-parity, Google optimizes its sites for Chrome, leading to additional confusion for both knowledgeable and lay users.

Fun fact: something similar once happened in production in the early days of an iPad-based point of sale startup.

In certain cases, a previously-entered zip code was interpreted as the number of cents to tip.

The tip field was off-screen

I wonder how much the [1] hiding of scrollbars and [2] UIs with excessive amounts of whitespace, increasing the need to scroll also contributed to you making this error.

We have trouble in our org with autofill on our New Customer forms. Depending on the browser it can be very difficult to remove the saved autocomplete values.

(How) did you resolve this?


They give you back $1000 if you do that?

The story might easily be worth that in marketing for Firefox.

I really want to know what the flagged comment said now.

You can see dead/flagged comments by going to your settings page and setting "showdead" to yes.

“Switched to Firefox.”


Is that allowed given that you appear to have authorized it and they have proof (it's obviously not fraud or something like that)?

Of course it's allowed. It would be under one of the Merchant Error chargeback codes since the intent was never there to tip $1000 (and it happened due to an error in their software).

If it's not classed under that code there's also 'Friendly Fraud' code for misrepresentation.

It’s allowed once, at the very least, if the merchant is hostile or unreachable.

Most credit card companies it would be no problem at all. It shouldn’t be necessary though, a quick call to the merchant should be all that is necessary for them to void it.

How do you pay with Chrome at a restaurant?

Probably for pickup or delivery.

As I had already commented on the issue, it completely breaks Germany's main train ticket selling website:


They have tagged the field as autocomplete=off but Chrome just doesn't care.

Also see this linked issue where they collected valid use cases for autocomplete=off. They just seem to ignore 452 use cases (I can't comment on the quality of them, I did not read any).


OT: direct link bypassing imgur’s horribly hostile UX https://i.imgur.com/BjYTgSn_d.jpg?maxwidth=1640&shape=&fidel...

Imo, valid use case for autocomplete=off is "the developer of webapp wants it".

Literally that and nothing more.

It's called "user agent", not "developer's agent". We'd be in a terrible situation if the browsers just followed developer's whims. Cf. popup blocking.

Browsers already follow developer whims, see: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Fe...

YouTubeTV uses feature policy to disable the browser's picture-in-picture feature.

Dismissing the above use cases as "developer's whims" is the fundamental issue most people here are taking with these decisions.

I think we can all agree that browser behavior should not be left solely up to the developer and is not a black and white issue. Nobody here is arguing that. We are arguing for following a guideline that makes sense. This is why we have the w3c, an organization that attempts to weigh the needs of user, developers, and browser maintainers.

The issue is that the browser is supposed to be the meeting place for negotiating between developer and user preferences. Its job is to take into account preferences of both sides, and render the site accordingly. Not to be a third party at the negotiating table. Breaking agreed standard in a way that can't be overridden by the user? Browsers should never do that.

> Breaking agreed standard in a way that can't be overridden by the user?

This was done by Microsoft by the old days, now it's Google. How ironic.

Correct behavior is debated and decided in public, resulting in a specification. In this case the HTML spec says "should", not "must". HTML Spec:

> When an element's autofill field name is "off", the user agent should not remember the control's data, and should not offer past values to the user.


> SHOULD: This word, or the adjective "RECOMMENDED", mean that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course.

As I read it, this behavior may or may not be a good idea, but it is not a violation of the standard.

Disclosure: I used to work on the Chrome team at Google, but I have no particular knowledge of autocomplete.

In this case, it can be overridden by the user, just not the developer :)

Not really if the user is non-technical and doesn't even know they'd have to override it.

"this website has asked us to NOT autofill your saved info.

if you would like to autofill the form anyway click"

i made it a little terse, but there has to be a way to make it succinct and human readable.

I think this could still be confusing as enough users will likely have no idea who "us" is in that message - if they understand the difference between browser and website at all. It could also be confusing if the website already provided its own autocompletion via JS: The user would get a message that autocomplete is turned off while they see that it seems to be right there.

But I think in general, some kind of prompt or override would work. I absolutely agree that if a user wants to use the browser autocomplete functionality, they should have an option to do so. I have no understanding for websites that just want to disable autocomplete without replacement.

However the concerns seemed to be about autocomplete being incorrect or conflicting with application-provided lists. I can see how that leads to frustration and confusion with users.

The Chrome team seems to trust its algorithm to an amount where they don't seem to find it necessary to deal with incorrect results - a view which doesn't match reality apparently.

Who is addressing the user here?

GP is explicitly not dismissing the above use cases, but merely the supposed justification of "the developer wants it" being enough.

> I think we can all agree that browser behavior should not be left solely up to the developer and is not a black and white issue. Nobody here is arguing that.

GGP was literally arguing that: https://news.ycombinator.com/item?id=21239172

Should developer be able to make it impossible to close browser or open 100 new tabs? No.

Should developer decide that fields are autocomplete off or green or show javascript warnings? Absolutely yes.

If user wish to change that, users thing. The browser/google has no business to be mediator here, second guess application they know nothing about and manipulate it.

The browser should be predictable, well specified and harmless. It should not force me to convert all ids into random strings just so that random data do not get prefilled in.

I've heard plenty of people on HN say password managers should ignore autocomplete=off and I agree with them. Because that setting is mostly applied by organisations like banks who incorrectly think they're making things more secure by doing so.

IMHO there are cases where autocomplete=off should be respected, and other times when it shouldn't be - it's certainly not as simple as saying always do or always don't respect it.

Password managers should ignore autocomplete=off in login screens, but not in administration screens where you’re editing other people’s credentials.

IMO the distinction could be made to not automatically fill when the autocomplete=off but instead add a button to let the user initiate it

>not in administration screens where you’re editing other people’s credentials

    <input type="password" name="other-users-password" id="other-users-password">...

That was recommended by Google 3 years ago, but there are a bunch of reports in the bug reporter thread that that doesn’t work now. And credentials go beyond passwords - names, usernames, ID numbers

Well, all major browsers now ignore it on password fields so the browser ship has sailed on that one:

> This is the behavior in Firefox (since version 38), Google Chrome (since 34), and Internet Explorer (since version 11). - MDN

I recently received a ticket from or security team to add "autocomplete=off" to all our password input fields.

I was pretty sure that it wouldn't do anything except make the security team feel better.

* should a developer be able to hijack browser shortcuts, no. Should a developer be able to disable right click functionality, no.

> Nobody here is arguing that.

Well, OP that I responded to literally argued that.

I agree the platform needs to expose useful features in a predictable manner for application developers.

But I'd much rather have browsers decide what's reasonable control over the user experience.

If the user installs extension to auto fill everything, it is on him.

When the browser decided to ignore spec, it is not user agency at all.

The need for auto fill is extremely application specific and the action is quite often destructive. And it is developer who gets to be blamed for lost data.

How do I install extensions into my most commonly used browser (my phone)?

Start using Firefox on your phone.

(on Android)

That's a bit of an appeal to extremes. Visually rendering static HTML per the spec is a situation where a browser "just followed developer's whims", and it's why protocols and standards exist.

That a browser could reliably infer from context what the right information to insert into a field could be is one of those things that gives great demo and may even work more often than not, but it's a virtually impossible problem to solve comprehensively. Developers have plenty of valid reasons to disable autocomplete and autofill functionality. Taking that away from them and requiring an obfuscating workaround is tantamount to creating an invisible pseudo-standard.

Imagine what would happen if we decided to solve the scourge of signed-unsigned pointer value bugs in C++ by having the compiler virally assume unsigned variable typing for memory related operations based on variable name. It might feel like a quick fix to a widespread problem, but it would break down all over the place and would break the code of developers who A) followed the spec B) knew what they were doing and C), at least some of the time, were doing it for a specific purpose.

These sorts of changes remove the incentive to write standards compliant code, leading to a mangled universe of hacks, quirks, and, in the end, bugs.

But I'm developing the application for the users.

It's pretty annoying that we have to make hacks for a client card page not to automatically assume you want to fill in your own details.

Then ask the user before filling anything.

And e.g. disabling Paste. But it seems clear that autocomplete gets it wrong often enough that developers should have some say in the matter.

Unfortunately a few developers are morons who misuse features, and browser vendors try hard to work around them. Case in point, lots of websites used to put `autocomplete="off"` on password boxes, which breaks some password managers. IIRC that’s why Chrome (and other browsers) decided to sometimes ignore the `autocomplete` attribute in the first place. Of course that doesn’t justify ignoring it completely (just for password fields) but maybe a similar reason is at play.

Exactly. Autocomplete=off gets misused. For example, there was a browsergame I played where because of idiotic "security considerations" autocomplete=off got applied to the login screen. At that time for me that meant typing in the password manually, thus picking a bad password.

So it's a good thing in that situation when the browser ignores the attribute.


No, it does not have its uses in those scenarios. Every website with login could be used in such a scenario, in an internet cafe for example. It's not on the web developer to predict this, it's on the terminal to prevent that issue.

It has its uses in some very limited scenarios, like a password field in a blog editor that when set locks the article for users not knowing the password. But that's not a login screen.

Me picking a bad password was absolutely caused by this misuse of autocomplete=off, it had everything to do with it. How can you claim otherwise? I was there, you were not.

Now they need to ban intercepting/blocking `onpaste` in form fields.

This has the added benefit of disabling local password managers like PasswordSafe. Might as well throw in some stupid password policy (like maximum password length, or banning some special characters but not explaining which) to make sure stubborn users won't use a secure password even if they're willing to type it in every time.

As an fyi, you can disable the onpaste event in the dev console. Obviously doesn't fix the actual problem but at least helps retyping your email...

That was even mandated for all password fields by the PCI scans required for any sites accepting credit cards.

What if the user doesn't?

It's the same meaning but I would prefer "just follow the specification an let developer decide what they want"

Finally i know why this happens only in Chromium. This irritates me for quite some time now...

Seems that Chromium based browsers aren't favorable any more: Tracking, Bugs, uBlock extension is flagged, Manifestv3, etc.

But Firefox has the same problem since it ships with Pocket and other sync stuff. I know that they really do care but they have problems of their own which really make me think which browser to use. The actual situation right now is very frustrating...

Pocket and sync are opt-in and seem pretty tame compared to the stuff in Chrome, just saying...

I use Firefox as my main browser on all my machines and haven't once been bothered about Pocket or other sync stuff. I vastly prefer it, for moral and technical reasons. I feared the switch from Chrome would be very hard and aggravating but it wasn't.

(The only thing that got me the first few weeks is that opening an incognito window is ctrl+shift+p instead of ctrl+shift+n. Once I got used to that I realized it actually makes sense because ctrl+shift+n re-opens a closed window just like ctrl+shift+t reopens a closed tab.)

ctrl+shift+p feels more natural to me as Forefox calls it "Private Browsing" whereas Incognito doesn't even start with a n

(Firefox user here) I find that ctrl+shift+n makes more sense: ctrl+n for a new window, add shift for a modified (incognito) new window. Although whichever shortcut is a very quick habit change to make anyway

Can anyone explain why the same behavior does not occur on https://www.sbb.ch/, the Swiss equivalent (which is based on the same software by HaCon)?

E: as soon as you submit a search on sbb.ch, the same problem occurs.

Did you try using a first-letter of a location you commonly use in forms? Here is what mine looks like when I use 'm' (like OP):


and here's what it looks like when I enter 's':


(actually, S does not autocomplete in the Swiss site)

You are correct! As soon as I submit a search query on sbb.ch, the same problem occurs with that query.

I always concluded that it was the fault of the website creators. Guess I have to switch to Firefox...

I think it might have that behavior because they have the auto-complete tag set to...

<input id="fromField" type="text" inputmode="text" class="mod_textfield_control" placeholder="Von" autocomplete="new-password"...

Though, there is also a lot of other stuff going on in that html element I don't understand.

That's an interesting hack as autocomplete=new-password will generally disable autocomplete.

Though it may not work every time, as browsers & password managers might also suggest new passwords (following whatever their internal generation rules are).

randomised field id works as well

does the escape key close the extra context menu?

Yeah! Good to know. Did the chrome UI make us aware of that shortcut at some point?

I discovered it during a panic once, never saw it mentioned anywhere.

Also occurs there. See https://imgur.com/a/4K73twK

In general, I don't like software defaulting to convoluted "grandpa-friendly" behavior instead of the simple, no-bullshit behavior. Moreover, if you want the simple, sane behavior you're a weirdo and need to justify yourself.

Another example of this is chrome hiding "trivial" parts of the URL. i.e. any subdomain that coincides with "www" or "m".

I'm guessing chrome ignores, it because people often place the option when they shouldn't. I don't understand why browsers can't simply respect setting and perhaps provide option for the user to modify behavior per site per field with ability to save. They could go even step further and do what Opera used to do i.e. provide same defaults for most popular sites.

I had this issue the other day, I solved it by generating a GUID each time the page loads and appending it onto the end of the ID.

Not ideal, but worked!

This has turned into a sad chicken-race between Google and developers, with lots of innovative workarounds on Stackoverflow.

Their tactic of overruling web developers doesn't work, it only make things more complicated for everyone, since many of the workarounds have other negative side-effects.


    ## Example 1

    For a reliable workaround, you can add this code to your layout page:

    <div style="display: none;">
     <input type="text" id="PreventChromeAutocomplete" 
      name="PreventChromeAutocomplete" autocomplete="address-level4" />

    Chrome respects autocomplete=off only when there is at least 
    one other input element in the form with any other autocomplete value.

    ## Example 2

    Simply make your input readonly, and on focus, remove it. This is a very 
    simple approach and browsers will not populate readonly inputs. 
    Therefore, this method is accepted and will never be overwritten by 
    future browser updates.

    <input type="text" onfocus="this.removeAttribute('readonly');" readonly />

    Style your input accordingly so that it does not look like a readonly input.

    ## Example 3

    Tell Chrome that this is a new password input and it won't provide 
    old ones as autocomplete suggestions:
    <input type="password" name="password" autocomplete="new-password">

I think it mentioned chrome is even now ignoring ‘display:none’ and ‘visibility:hidden’ declarations so those workarounds no longer work either.

Great! :( Now I need to recheck to see if an autocomplete bug has reappeared where it thinks an email address field is a username field in a change password UI and autocompletes the email address with the username from the login page.


I feel your pain - sorry you have to deal with it. I’m working on a client data management app and I’m having issues everywhere with this. It’s been an open ticket on Jira for months!

Why would they ever do that. That’s just asking for autofill phishing.

Preach. Them changing the rules and overriding the usage of this attribute is ridiculously stupid in the first place. HTML attributes are there for a reason, developers are supposed to be able to use them and they are supposed to work as expected. But in order to push their password management and form filling functionality, they just give everyone a finger.

At the same time many many many dumb corp-o-rat shit sites disable autocomplete because cOmpLiEncE. And users love autocomplete...

Naturally the problem is choice. Chrome/G should show a button to auto fill fields.

They could simply unleash some AI magic, train a network to recognize good and bad sites, etc. (Aka. the Apple way. As they also have a we know better policy, and users seem to love that.)

I haven't tried it, but maybe div with "contenteditable" could help?

This is what I resorted to on one of our internal tools that wants a phone number to be entered and kept getting auto-completed with our staff'S phone numbers.

I wouldn't recommend it for public use. There's a lot of subtle behaviour around an input field that you'd need to replicate manually.


Because Chrome has 70% of the global market share, and web developers should know what their customers use?

When IE had a huge market share I would use and build with chrome then test and correct in IE. It’s possible to support clients while not personally supporting this.

Developers need to understand this is war and act accordingly.

This is not "war". Don't overdramatize.

Don't be toxic and rude.

Someone disagreeing with you is not "toxic".

> Don't overdramatize

Is toxic.

it's literally what you were doing

> “ This has turned into a sad chicken-race...”

great dash use! i almost read that as a race of sad chickens (seriously) but the dash autocorrected me. bravo!

Setting aside the merits/lack-thereof of this particular decision, Chromium ignoring established web standards like this is especially dangerous as we're trending towards a world where 1) Chromium itself powers the most popular browser in the world by an increasingly unhealthy margin, and 2) even competing browsers are increasingly becoming skins on top of Chromium.

We are becoming more and more reliant on the developers of Chromium to be steadfast stewards of the standardization process. Their massive influence means that any deviation from actual web standards on their part will inevitably create a new and conflicting de-facto standard that will create decades of lasting damage and irreversible tech debt for the entire web (eventually leading to a repeat of the IE6 dark ages).

Decisions like this demonstrate an utter disregard for the crucial role Chromium plays in the web standardization process, and jeopardizes the entire ecosystem.

W.r.t. this particular decision, I feel a more reasonable approach here might have been to allow users to manually trigger autocomplete on individual fields with `autocomplete="off"`, and/or to allow users to strip away the ability to disable autocomplete on a site by site basis.

Analogously to how users are able to selectively grant access to certain default-off features for each website (notifications, camera access), users should also be able to revoke access to certain default-on features for sites that abuse them without affecting the vast majority of sites that use the features for their intended purposes to create a better user experience.

This isn't even the first time we've seen this either. Both Chrome Mobile and Safari Mobile go against the standard and implement `vh` incorrectly.

Honestly it seems strange that they go against the standard though considering how much power they have in defining it. Why break from the standard when you can just update the standard. It ends up being the worst of both worlds - documentation that says one thing (that they had a hand in building) and an implementation that does something completely different.

Admittedly, the viewport units are just all-round badly thought out and fundamentally broken. (The corresponding problem with vw is that the viewport units includes viewport scrollbars, so that on a page with vertical scrolling, `width: 100vw` will cause horizontal scrolling on platforms where the scrollbars take space.)

Genuinely curious but how are they fundamentally broken? Or at least, any more so than px or cm or em etc? They all have weird edge cases at extreme usage scenarios... Or rather, if you were to 'fix' vw/vh, how would you go about it?

There are two fundamental problems.

First is the scrollbar thing I mentioned. It means that you can’t use it for layout purposes; you thought you could sit four blocks side-by-side with `width: 25vw`? (And that was pretty much the whole reason why people wanted viewport units in the first place—this being before flexbox provided alternatives that are generally acceptable, though not without flaws, or grid provided often better alternatives.) Sorry, that’ll only work on a platform with overlay scrollbars, not where scrollbars actually take up space. There’s fun history around Firefox’s implementation where they made it possible to get the “right” behaviour, but no one else implemented that, so it was eventually removed from the spec.

Secondly, it was also predicated on the idea that the physical viewport size won’t be changing all the time; but on mobile platforms where the address bar can get out of the way as you scroll that’s simply not true, so you’re stuck with browsers having to decide between two interpretations: vh picking either the smallest or largest value and thus not actually reflecting the viewport height half the time, or vh reflecting the viewport height and causing the page layout to jump around in a jarring fashion when the address bar expands or collapses, when vh is used as part of the layout.

The first can be fixed in two ways (which can be combined): firstly, by defining new units that exclude document element scrollbars (e.g. 100vw2 ≅ calc(100vw - env(scrollbar-layout-width-auto, 0px)), to define it in terms of what follows); secondly, by exposing the layout width of scrollbars to CSS, e.g. as env(scrollbar-layout-width-auto) for the width with `scrollbar-width: auto` and env(scrollbar-layout-width-thin) for the width with `scrollbar-width: thin`. These values would be something like 17px and 8px on Firefox on Windows, and 0px and 0px on platforms with overlay scrollbars.

The second, you probably need to expose new constants for the possible extreme values of viewport height. e.g. I could imagine env(viewport-min-height, 100vh) and env(viewport-max-height, 100vh) working, which would then allow developers to select one or the other, as suited their purpose. Or define new units opinionated on which value should be used, and then deprecate vw and vh since they’re inconsistently implemented.

(In using env(), I must caution that it’s currently rather broken for some sorts of situations: https://github.com/w3c/csswg-drafts/issues/3285.)

All browsers on iOS are Safari wrappers

In this case I'm referring to Chrome Mobile on Android.

Strictly speaking, Chrome is not ignoring a web standard, since the standard does not require this behavior (no "MUST" keyword).

Well they are ignoring hundreds if not thousands of developers which is the main issue at this point.

Well, that's what Chrome always does? Just like the preventDefault breaking change thing. They have no problem ignoring developers if they think it will be a net benefit for user experience.

Yup. Very frustrating as a dev. I don’t really get how they rationalize it either - having their autofill appear over and obstruct a developers autocomplete functionality is not user friendly!

By ignoring "hundreds if not thousands of developers" they are respecting the wishes of millions of end-users that don't want the site owner to decide what they can and can't autofill.

Obviously a simple boolean is the wrong design here. But can you suggest a better one?

I don't think "millions of end-users" want broken websites because now another developer (Google) gets to decide what's correct (instead of the web dev).

A better solution could be to just behave reasonably by default but allow the user to re-enable auto-fill with a single click.

Surely there’s a way to check if a field has some kind of programatic interaction like typeahead. Maybe check for that before taking over with autofill? I don’t know - maybe that would break spec or something - I’m not a browser developer. I just know the pain of autocomplete=“off” not working as expected.

That depends entirely on your interpretation of RFC 2119 in this context – whether you consider Google to have "valid reasons" or not. It's not an objective or inarguable issue.

Point taken. Maybe in this case the spec authors were also at fault for leaving too much leeway in the implementation.

Nevertheless, from their stance on the issue so far it stands to reason that the Chromium team would have objected to any efforts to tighten the spec, which would also have led us to this same situation.


I think you're stumbling over too many negations here. A single browser running the web is a bad idea. Diversity in browsers is a good idea. If any statement related to this could be heresy here, then it would be heresy to suggest that a single browser running the web is a good idea.

HN comments pretty consistently argue that browser monoculture is a problem. I don't know where you're getting this.

Because other people here are throwing in their frustrations, I will at least add that on the flip side I have been frustrated by sites that attempt to disable autofill for illegitimate reasons, like attempting to disallow password managers. I think I understand where this is coming from.

On the other hand, I, too, have been bit by this at least once, in the past. I think it was easier to just disable it at that time, but IIRC, the solution we landed on was to not use form controls at all but switch said text controls over to use content-editable elements. In our case it made sense, since it was not a form at all. My memory could be a bit hazy here, though.

(I do not work on Chrome or use Chrome at home, but as disclosure I do currently work at Google.)

Chrome explains in their security FAQ [0] why they don't adhere to autocomplete=off for password fields.

They could still follow it for other fields.

0: https://chromium.googlesource.com/chromium/src/+/master/docs...

Of course then websites will just stop using password fields and that is awful for security and convinced.

Shouldn't the work on changing the spec, rather then confusing majority of developers and causing even financial loss and embarrassment for many? This thread is filled with cases where this caused issues.

Google should spend their cycles to work on a standard for password managers. The current design of those is flawed anyway.

I think we need a way to disable features only for those developers that abuse them. Like uMatrix but built-in and with rules being supplied automatically as ad blocking lists are.

You autocomplete=off a password field? That attribute won't have an effect on your site anymore.

You auto-play videos when the user doesn't expect it? What videos? The web doesn't support videos – as far as you are concerned.

Scroll hijacking? I hope you used progressive enhancement because you just lost your Javascript privileges.

This is actually a reasonable proposal. A moderate list could be an in-built feature, with some config options.

It would only work if a strict list is the default in browsers that developers are not willing not to support. The point is not to hack around user-hostile patterns. That frequently means entering an arms race you cannot win. Even in the best case, you're doing that site developers' job. Badly: You have to remove the bad stuff while avoiding overblocking. That's a lot harder than not adding it to start with.

I'm proposing to take away tools from developers that abuse them no matter how much that breaks. If you mine bitcoins or annoy the user using pop-ups we shouldn't spend time finding out what scripts are good and which are bad but just disable all of them, till you remove the user-hostile stuff. We just broke your SPA? That's your problem. Better start working then.

I asked if something similar was coming to Brave, and the answer was yes. Opera maintained what i believe was a somewhat community powered site patching repository back in the Presto days.


Can we have a reliable cross-browser way to say "this is a change password field, so don't autocomplete it" and "this is an email address field not a username field, so don't autocomplete it with the login username"?

The autocomplete="new-password" (or autocomplete="off" or even autocomplete="some-random-nonsense") method is supposed to be doing that according to the spec. It is ignored by browser developers probably due to abuse. Additionally, all of those also trigger the prompt to save the password. For a password change form, that is the right thing to do, for many others it is not.

I'd very much like to see a way to disable that save prompt too, especially in user management screens. When I create a new user, I want neither my own password to be autofilled, nor do I want to overwrite my saved password with theirs.

The web needs strong typing. Right now most input fields are variants of text boxes and there are no naming standards that can be enforced without breaking millions of sites.

I've seen/used autocomplete="new-password"

Developers can say that. What we need is a reliable way to know if they're lying.

I even found a name for such an extension: PunishIt.

Disabling autocomplete is so incredibly frustrating that I ran an extension in Safari to remove the off tag from sites. Of course I want to use KeyChain, the whole point is that touch based ID is more secure.

I have my own extension for Chrome that yanks the autocomplete attribute. I much rather deal with over eager completion than websites disabling it for silly reasons and pseudo security.

Touch based are not better. Your fingerprint is not a password, it's just an identifier and shouldn't be treated as a secret.

Your fingerprint is an additional factor (something you are) for the actual better-than-passwords thing here, the isolates processor/storage in your laptop that handles Touch ID (something you have). Without physical possession of the laptop, the fingerprint is useless.

> Your fingerprint is not a password

Correct, because a fingerprint makes a password to some extent redundant.

> it's just an identifier and shouldn't be treated as a secret

Correct, identifiers are not secrets. Your face is not a secret and your fingerprint either. The problem is that we use secrets to identfy someone, when we potentially already have tech which can identify someone without having to remember a secret and store it in a dictionary of secrets on someone else's computer in the cloud.

The sole purpose of a password is to identify someone with a certain degree of confidence. If a fingerprint taken from a handheld device, which has already been proven to belong to a person, can provide the same if not even a higher level of certainty about someone's identy, then a password or as you say "secret" is not required at all anymore.

Um no. The user Name or ID is to identify.

The password is to secure that identity. With TouchID or FaceID you are using them for both. Which reduces security.

> then a password or as you say "secret" is not required at all anymore

Definitely incorrect. Someone can cut your finger off, lift a print off your coffee mug, extract it from a selfie etc. There's been dozens of ways to exploit over the years, many of which have hit HN.

What is the compromise? What is insecure about me being me? Either we talk past each other not understanding each other's point, or I feel like you profoundly misunderstand the reason for people using username/password for authentication nowadays?

With most things you need to authenticate to gain access. Authentication is only trying to solve one question - identifying that a person is who they say they are. If I walk home and my wife sees me, she can identify me immediately by simply seeing my face and other biological attributes which gives her 100% certainty that I am who I am, therefore she doesn't question me on entering the house or calls the police.

If I was to go through some top secret lab experiment which would change my look (make me younger by 10 years or something) then I'd struggle to walk home and convince my wife that I am me without providing additional evidence, like sharing some secrets which only she and I know and we know that nobody else would know.

With technology so far we didn't have the ability to identify someone as confidently and as fast via biometrics (like my wife does) as via other means, which is why a few decades ago we had to invent a workaround, namely username and password - which is a secret that hopefully only me and the website knows. This was to date the best way to identify someone, but times are changing fast, as as biometric identificaiton via technology is advancing fast, we are more and more removing the need of username + password for authentication.

Hope now my point makes sense.

A biometric identifies you. An identity does not inherently authenticate you.

Identification is knowing which account to log in. dustinmoris, User123 or user124.

That alone is useless, as anyone who knows your name could log in. So we need to add security to authenticate you. To authenticate with reasonable certainty that the person accessing dustinmoris actually is Dustin Moris, at this moment willingly accessing their own account, willingly transferring £1m to NeedMoreTea. :)

Now, it's certainly true that biometrics are uniquely associated with you, but prove identity not authentication - bear with me here.

You leak DNA and fingerprints all over the place, constantly. Fingerprints have even been picked up from photos. That makes fingerprints surprisingly weak authentication. Face IDs have been fooled by video and photos. I'm not at all current on what the state of the art on FaceID hacking and devices is, so can't say more there. In a police interview they can place your thumb on the phone or wave phone under your face. Maybe borrow your thumb while you sleep to transfer that £1m. The bank will say you fully authenticated it, in their highly secure app, you will deny it.

In combining biometrics as identification and authentication we've compromised the system. It does not give certainty, or even especially high confidence. If motivated, anyone in your office could lift a print while you're out. Computer or phone says Dustin Moris is believed awake and accessing, but you're certainly not doing so willingly, nor can you resist if just a touch or look unlocks. In this context, your wife might easily pick up on whether you were willing in a police interview or not, the phone can't, because mere identification is the unlock.

With a password they have to persuade you to reveal it. It is very easy to not leak passwords accidentally on PostIts, and manage the complex with a password manager. The police can certainly beat it out of you, but most places are supposed to have rules against that. Many places actually honour those rules. A judge might question the bruises and missing teeth.

For a lot of people that's a fine distinction that doesn't matter, the convenience of it probably being you willingly unlocking the phone with your thumb is good enough. If security actually matters, it's not.

Sorry that got a bit long.

You highlight flaws in faceid, etc but passwords also have flaws.

- Fingerprints can be lifted from pictures. - Passwords can be forgotten and thus have seriously flawed reset schemes that often fallback to something as simple as having the right phone or backup code. - passwords can be lifted by keyloggers - passwords can easily be phished - passwords can be shared

Authenticating access can come from 1 or more of: something you know, something you have, something you are.

They each have flaws. They all do the same thing.

They don't do the same thing. All the flaws in passwords can be avoided, the semi-public state of biometrics cannot.

Authenticating access, if you follow standard security practice, should be at least two of, and preferably all three of: something you know, something you have, something you are. Not one or more.

There is no axiom that the flaws of biometrics cannot be avoided. “Something you know” is in fact a biometric property of the brain that can easily be transferred.

Touch ID combines something you have with something you are. It’s easily better for 99.9% of the threats out there that result from weak passwords shares across sites and users trained to type them into anything that resembles the real logon page.

If we had technology that doesn't exist, we could have secure biometrics.

In practice the computer is just getting a picture, which is semi-public information, and that's why the only thing you can rely on is that it's an identifier. It's not enough to authenticate when you really need security, only in more casual situations.


How many emails require a password to send to them? None. It's identifier. The password is the security.

Using biometrics - an identifier - for security as well is a compromise for convenience. As you say yourself biometrics make you who you are, they don't secure who you are.

> The password ... is a way of identifying that you are you.

That is not correct. The "system" identifies you by your ID (username - biometrics - whatever). And for the system to ensure that whatever actions you perform on it, are really coming from you, you have to add proof of your identity to the orders for those actions. You use a secret to create that proof of identity. The secret is yours, the identity is the system's (what it uses to identify you).

If you use something you don't treat as a secret (such as your fingerprint) as your proof of identity, you'll be loosing it left and right - you'll be inviting everyone around to impersonate you. Drinking a can of coke and throwing it to the bin would be the same as printing your password in cards and passing them around.

Anonymity is weakened if we tie authentication to biometrics. Something you know (as in password) is always theoretically more secure than something you are (your physical characteristics).

I think the statement you make about "something you know" always being more secure is not nearly as clear cut as you present it.

A combination of "something you know" and "something you have" is always going to be a very strong authentication scenario, and making "something you have" a non-hackable thing that truly only you can have (i.e not a USB key or a TOTP seed etc.) is a good choice.

The catch here is that when you mention biometrics, you make the assumption of static biometrics (rightfully so, as most methods like Touch ID and Face ID are static), but if you combine lets say face biometrics with liveness checks, you are getting into a territory where faking them becomes much much more difficult (there are various mechanisms out there, the good ones rely on completely random interactions and light-bouncing detection methods as an example).

The real challenge is how do you make these very strong biometric methods frictionless and cheap? Or how do you introduce similar controls - like liveness - for easier methods like fingerprints?

And using any of these (ideally) has absolutely nothing to do with anonymity. You are not anonymous online regardless of what you use for authentication - and you are frankly not smart if you assume thats the case. A company offering a service with biometrics is no different from a company doing the same based on email/username and a password, if they do privacy and security right.

I could actually get into a much longer rant about this last part, as it blows my mind how many netizens are all about privacy and whatnot, yet are willing to expose every single detail about them when its convenient from them...

> And using any of these (ideally) has absolutely nothing to do with anonymity. You are not anonymous online regardless of what you use for authentication - and you are frankly not smart if you assume thats the case. A company offering a service with biometrics is no different from a company doing the same based on email/username and a password, if they do privacy and security right.

Why should I use unchangeable, personally identifiable details when logging into some random joe's website? The mechanics of it are also tied to specific devices if I'm not wrong. I can't just login anywhere without risking leaking secrets. When did fingerprints, iris patterns and DNA go from necessary tools in law enforcement and biology (and for highly secure installations) into casual usage all over the place?

You don't have to identify yourself with a "random Joe's website". You can read the comments here without identification, you can seach for flights without identifying yourself, you can search for hotels without giving away who you are, you can read the news, watch YouTube, etc.

But when you want to access something that should only belong to you, e.g. your bank account, your Google Drive files, your private emails, etc. then you must find a way of identifying yourself with given website. Username/password is one way, biometrics is another way.

Only my family has permission to enter my house. When someone enters my house who doesn't look like my family, then they can throw around all secrets, passwords and usernames they want, I'll kick them out, because in my eyes they don't pass the ultimate test of verification, namely biometrics.

So far we were unable to provide the same level of identificaiton via the web, but technology is changing rapidly, so I don't see why username/password are always going to be more secure. On a theoretical basis it doesn't make sense. Because inforation can be easily shared, spread and copied. My physical composition not.

Biometrics are digital interpretations, which can be copied. The biological sources cannot be easily changed. So _when_ the digital copies leak they become a liability.

Passwords can be easily changed. And until reliable, remote, mind reading technology becomes available they'll continue to be better than biometrics.

> Anonymity is weakened if we tie authentication to biometrics.

I think you mean pseudonimity. There should be no authentication in anonimity.

> Something you know (as in password) is always theoretically more secure than something you are (your physical characteristics).

If you authenticate, you can use methods that are more or less secure. If you use a password method, and use a secret as password, it will be more secure than if you use a biometric as password.

But the security of the authentication method will bear no weight on the "pseudonimization" - meaning on the difficulty of linking the authenticated identity with your real, legal identity.

It's effectively like using one password for every app/site authentication. What if the hashes/keys leak out? Then how easy would it be to change your keys for other sites while still using the same biometrics?

Anonymity is of course weakened, but that is a different debate. Currently there is no anonymity at all if your account is linked to your email address, which you've also used to register your online banking account, your credit card verification, your PayPal account, etc. and when you have your mobile phone number confirmed, etc.

A password gives you 0 extra anonymity in this case. All it is used if for "identifycation" and in this regard it sucks in comparison to true biologial identification.

You are mixing up terminology. When logging in a website, your username identifies you, and your password authenticates you.

Password authenticated key exchange algorithms (PAKE) solves the dictionary of secrets problem, by hiding the password from the server.

We recently had to go through a website project and disable autocomplete on all password fields after a security scan as part of a PCI compliance process. For autofill password it didn’t seem to have any effect in _any_ browser I tested in. How can it be used to thwart password managers?

Why would you need to disable autocomplete passwords for PCI compliance?

Because either the PCI standard says so, or the PCI consultant says so.

PCI compliance is about checking boxes, not weighing the options and making good choices.

Exactly this. Some of the checks _are_ valuable. We found a couple of real issues and made good security improvements. But we also ticked off more than a few boxes that made no damn real sense.

I ran into this last week (with LastPass, not Chrome - this seems to be a common practice):

I have a form where users enter information about their suppliers (I make restaurant management software). This includes a field for the contact email address, which LastPass was autofilling the email address the user used to log in. This happened silently, quickly enough that users wouldn't notice it on page transition, and would overwrite the initial value. Even with autocomplete=off.

This was a DATA LOSS bug - users would load the page, make a few changes, save, and not notice that they'd lost the email address they'd stored for the supplier. Fortunately LP has a method to force disabling of autofill (data-lpignore=on), but this could have all been avoided if they'd followed the spec I was relying on. I still don't know if some other password manager, maybe built in to Firefox or something, will make the same mistake, haven't had time to check yet.

Similar thing at my last job. Some internal webapp that was communicating with a bunch of other services. Whenever you edited the connection to such a backend you'd get a bunch of settings to change, as well as - depending on the backend type - credentials to talk to that other service. So if you were saving your credentials to that web app, whenever you edited such a backend connection, it would overwrite the username and password field in that settings page with your credentials for the web app itself. Even though the fields were pre-filled with the current values in the html the server sent you, chrome just went ahead and replaced them on page load. It happened more than once someone accidentally saved the settings and replaced the actual credentials for the backend with their login credentials for the webapp. So anyone accessing that edit page afterwards could steal their coworkers password.

So at some point a colleague went ahead, found the template and added a fake username and password field at the very beginning of the form that had something like "position: absolute; top: -2000" (after an hour of failed attempts with the autofill attribute and using hidden fields and whatnot).

So yes, f*ck those Google devs on their high horse.

The LastPass on is especially horrific because it fills in fields that are already filled, and it fires a change event. So if you're auto-saving on a change event, that data is lost as soon as the page is loaded.

The entire design of password managers that hijacking the DOM is flawed.

A recent LastPass version does not fail gracefully and a lot of my webapps started getting weird console errors from users with LastPass extensions that were trying to unsuccessfully inject into our forms. :/

As a user, Chrome's autocomplete went down the drain for me once they started to fill all fields at once. I tried to autocomplete one field and often chrome filled the other fields with unfitting data. This happened so often that I started to manually complete fields even if autocomplete was available. I still miss the good old times[tm] of single-field autocomplete...

(And for the record: As a developer, I have been bitten by Chrome ignoring autocomplete="off" as well.)

Yup, I like how FF does it. I basically just do a repeated sequence of:

TAB -> Arrow Down (select one of the previously filled values I've used) -> Enter

When filling a form. Given that sometimes we order things on made up names, it's quite useful no to have that autofilled when we're not wanting to.

This is a bit of a missing perspective. I wouldn't even care about them not following autocomplete=off if the damn autocomplete was useful. Instead 90% of the time I'm fighting it.

They really, really need to stop trying to autocomplete entire forms, and probably not anything other than "user/password" forms. The whole 'look at this input id and match it to a bunch of categories' approach has been thoroughly debunked as useless.

I guess this will lead to a horrible coding style where instead of having this in the form:

    <input name="email">
We will see stuff like this:

    <input name="16fkr9547kancot944128sddfksdf934998aafccugt75">
Where developers use some type of abstraction that generates a random id for each field and then assigns it to the original value server side or in javascript.

Just like they already randomise asset filenames to avoid caching.

I’m doing this with a field where I provide server side autocomplete of an item. It sucks, but I can’t find a better workaround.

MDN literally mentions jquery.disableAutoFill which basically does that for you under the cover (scrambling the name on display, and unscrambling on submission).

Your comment about randomizing filenames got me thinking. It would be great if we could add a cache key to the HTML element, that way we could cache everything forever with the same filenames and still invalidate things by just making the key being the SHA of the deploy. Too bad that's not a thing.

Subresource integrity is for ensuring that you got the right file (it fails if you didn't), it doesn't do anything with the cache, unfortunately.


Seems like it would be crazy not to use it for caching.

The problems are 1) side channels. Caching those would mean that a malicious script could use timing to find out if the user accessed specific other websites. Also it enables a channel for cross-domain communication. 2) malicious user tracking would not work as well anymore, which would be good but Google probably will not support that. Currently they get all those nice http log entries from sites that only include fonts etc.

For sensitive files with known names, you should be able to set that the cache is only valid for requests from the same host website (as the server might decide to send another response when the host is different). That would fix the sidechannel.

It's like Internet Explorer 6 all over again. SAD.

lol, years and years ago (like 2005) as an attempt at stopping XSS and CSRF attacks and bots I came up with a system that named all the inputs a salted MD5 of the intended name with the salt randomly generated then stored in the users server side session.

It’s still a reasonably effective solution for CSRF, though there are much simpler options, but today’s bots largely have cookie jars so you will likely need a CAPTCHA.

How would it stop XSS? Is it meant to stop attackers from getting JS execution on your site, or to mitigate the attackers' abilities after they have JS execution? I don't see how it would do either.

I mean both CSRF and XSS. If your inputs are named something different on everyone’s machine, per session, you can only XSS yourself. Not very useful, you can already do in many other ways.

Also my comment above is not an endorsement of the methodology, it’s just a thing that happened.

Why can you only XSS yourself? If an attacker has javascript execution on the website, the javascript can identify forms not just via their names, but by relation to other elements in the DOM tree. The javascript can for example do

document.querySelector("form > input");

That doesn't rely on names at all. Just like a human can use eyes to find the right input, the javascript can use the DOM tree to find the right input.

I understand that it provides CSRF protection. It's basically a CSRF token embedded in the name.

You misunderstand. Specifically on sites without cross-user visible content, the JavaScript needs to get into the system via an input, POST or GET. When every user input is named something different, you have no ability to send a link with a JavaScript payload.

Hmm, I guess it could provide some protection against some types of XSS, but it's not comprehensive. It doesn't provide any protection against stored XSS. And even for reflected XSS, it assumes the only dynamic thing in your URL is form inputs.

An example of something dynamic in your URL that's not a form input would be links to various articles posted through some type of CMS. For example http://example.com/article?id=foo or even http://example.com/article/foo . In both cases if foo doesn't exist, a badly-written error page could print out "Error, foo was not found" without escaping foo.

Maybe they mean CSRF? Then things make more sense since you can’t guess the field names cross-origin.

> Just like they already randomise asset filenames to avoid caching.

But why? Isn't the whole point of caching to improve delivery if static assets?

Did they run into staleness problems? Then why not use if-modified-since/if-none-match?

You are assuming browsers(there are many more than chrome&ff) ,proxies, etc. do caching correctly.

And you are assuming web devs could configure their servers properly. Anyway, it all boils down to the mentality of lowest common denominator: using that workaround, nobody has any incentive to fix their proxy.

Yep. We recently noticed this problem after renaming form fields for the purpose of accessibility (screen readers)

Autofill for offscreen elements gives me the creeps even without the data getting misinterpreted

This is why form elements should never be hidden after loading. Display none should be the default in the HTML for non-relevant content, which should be enough for most autofillers. It also prevents flashing of content when autofillers try to populate it, causing the hiding to delay, which I recently saw in a production app.

Frameworks like React and Vue don’t even render the HTML into the DOM until conditions are met so the situation is improving.

Another positive step away from jQuery hackery!

If you want to be malicious and capture wrong auto fill data, wouldn't a simple AJAX request on change be enough to capture sensitive data before the user gets a chance to correct it?

Yes, and there are popular third party analytics platforms that record all activity on a page, which includes not just every key press, but the speed at which it was done, with or without a submission of the form.

Web browsers should never, ever use auto fill unless the user has already entered that information on that domain already. Popular domains that host third party content should never be able to auto fill.

I'm not really sure but I think the behavior is a bit different with autofill. While the browser show the field highlighted, the data is there for _the user_ but the actual form field has not been actually changed and no dom event is fired.

If you think late rendering or "display:none" will put Chrome off, think again! The anvil.works IDE is Angular, so it's all rendered post-loading, and Chrome still autofills the weirdest things. Relentlessly.

Search boxes? Sure! App titles? Yeah, one week a bunch of our users' apps got renamed to the author's email address.

Being an IDE, of course, there are lots of places where inserting random text will break things. When I open up the console and see it filled with "'meredydd@anvil.works' is not a valid Python identifier"? That's the signal to go hunting for which dynamically-generated off-screen text box Chrome is stuffing credentials into this week.

TL;DR Chrome's autofill is out there. It can't be bargained with. It can't be reasoned with. It does not feel pity or remorse. And it will not stop.

That begs the question, at which point does autofill happen in an iframe? So I pay for an ad, and have a password, creditcard number, address etc. form in the background. Does the browser autofil, or does it autofill when the user starts to fill in a form in the foreground?

Asking for a friend.


When a phrase is used the “wrong” way more often than the right way, it ceases to be wrong.

What about "I could care less"? Seems like that one is just always wrong.

Thinking about language usage as "wrong" or "right" might not be the most helpful way to think about it: https://stancarey.wordpress.com/2010/02/16/descriptivism-vs-...

(p.s. not saying there's no wrong or right when it comes to ideas themselves)

Yeah. But people say it enough I've started mentally converting it to:

"I _could_ care less (but I can't be bothered)."

which is for all intents and purposes is more insulting than not being able to care less.

I think you mean all intensive purposes </s>


Well, that's one way it happens. The language is just the set of conventions the speakers understand. When those conventions change, through ignorance or any other reason, then the language changes. That doesn't mean you can't grumble about it as it happens, though.

I used to really, really agree with your line of thinking, and get an almost depressing feeling that this happens. But if you sit back and think about it, a lot of our 'correct' language started off as mistakes. Mayday, Ammunition, Culprit, to name 3 of a list of likely thousands...

Given that ignorance merely means not knowing, yes. Yes, that’s exactly how it works.

Unfortunately, yes.

I still have to come to terms with it.

Idiocracy, defined.

Before you say that, ask yourself which use is actually using the word "beg" correctly.

“Idiocracy” doesn’t appear in my dictionary, how did you come to know what it means?

Someone made a movie and called it that. It's not really a word. So yes, you proved words can be created. That however is an opposite process to the one that destroys the meaning of an existing word or expression through repetition by people who don't know how to use it. I'm against that kind of half-assery. Creation, art, those take work and discipline. And as much as creativity demands going against convention, you always first have to learn & be aware of what the conventions are. Not even doing that work, and then just claiming your way is just as legitimate as the accepted way for your own mere convenience, well it's a shortcut, and I mean in the best case you have a very slim chance of "just happening" to be a great artist, and in the worst case it's the height of ignorance and narcissism. So yeah I'm unimpressed. I would accept & respect even a deliberate act of transgression -- knowing the right way and doing it wrong on purpose -- before that kind of shoddy relativism.

Yes it does.

It's okay to say something with the same words as an idiom, without using the idiom.

The idiom was a mistake/mistranslation in the first place.

You're trying to push people away from the correct use of the word 'begs' into an incorrect use just because it's tradition.

Technically correct, but look deeper: Why is this mistake prevalent? What are people trying to say that causes them to borrow (misuse) another phrase to express it? This indicates a "hole" in the language that is filled with the wrong item.

Like a code review, do better than just saying "This line is wrong". Suggest an alternative. Especially one that flows naturally into the prose.

I don't see a hole. The correct phrase is "raises the question". People just confuse them because they're similar.

And worse part is that autofill still works in incognito.

Oh yes. I seem to recall it was demonstrated as a way to covertly steal user information from the browser a while ago.

Haven't browsers already done changes to combat that?

Curious as to the global business impact this has had. I personally have spent at least a dozen hours debugging forms and trying to disable autocomplete/autofill on my kiosk-based applications. How many development hours collectively have been wasted on this unilateral decision.

I have not been this frustrated since the days of writing css for ie6, and at least back then the devs response was more "sorry its our rendering engine" and not battre just saying GFY seemingly out of disdain.

I try to be as free-market as possible but I sure wish that the w3c had some teeth when it came to things like this.

You're not the only one. I remember spending a good two days on it this year for a page where the user puts in their credentials for external integrations.

Btw, at http://google.com, this is what they have for the search field:

<input class="..." maxlength="2048" name="q" type="text" jsaction="..." aria-autocomplete="both" aria-haspopup="false" autocapitalize="off" autocomplete="off" autocorrect="off" role="combobox" spellcheck="false" title="Search" value="" aria-label="Search" data-ved="...">

Does anybody have a clue why it does not show up there? The Google team must have done something in order to circumvent what the Chrome team did. Horrible.

Overall, I still believe that neither of the extreme strategies ("always honor autocomplete=off"

Is it “extreme” now for a computer to do what the user wants and not what a random Google employee wants? How does this differ from malware?

It's also in the spec.

People might not like the spec or it might be incomplete, but adhering to it is a very important part of improving it until it's a good one.

Now I'm no webdev, but I could very well imagine that the spec is already a good one. So the situation might be even worse.

No, it's not a requirement in the spec. Chrome is actually spec compliant regarding this issue.

The spec only says "should", not "must". Apparently these wingnuts thought that means the spec can be ignored.

> The spec only says "should", not "must". Apparently these wingnuts thought that means the spec can be ignored.

Not arguing in favor of this particular choice, but yes. That is exactly what "should" means in most cases.

For instance, in RFC 2119: https://tools.ietf.org/html/rfc2119

> SHOULD This word, or the adjective "RECOMMENDED", mean that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course.

Blanket disregard of a SHOULD directive as a UX decision neither falls under "in particular circumstances" nor indicates that they've "understood and carefully weighed" the consequences.

There are no semantics to discuss. Chrome is spec compliant. That's a fact. Words like "should" and "must" have well defined meaning that is clarified in every spec document, which I would highly advise you to read before writing any further comments and insults.

SHOULD does not equal MAY

autocomplete=off is set by website developers, not users.

However the more important point made by the OP still remains which was some random Google employee chooses to ignore it.

The way I see it, autofill off should mean off and if I click the icon in the address bar to bring up, well, let’s call it “quick access to per-site settings” then maybe I could override it for a specific page load or site, like you can Flash, etc.? This per-site configuration is starting to become “normal” given iOS 13 does the same in Safari for permissions, content blocking, automatic reader mode, etc. It would make sense that the default is “follow the HTML5 spec” but you could put a notice in the address bar if you really felt otherwise...?

Safari does a much better job by only filling visible form fields, I think (though it too has a tendency to put my address both in Line 1 and Line 3, which is annoying...).

Explanation from the 'rogue Chromium dev' is linked to in comment 19 of this bug: https://bugs.chromium.org/p/chromium/issues/detail?id=914451...


...which really makes it worse:

1. Our programming language has an attribute called "autocomplete" with two possible values: "on" and "off"

2. We will now (without consultation or announcement) simply start ignoring one of those values when you specify it. (and certainly not document the new behaviour!)

3. Here, I made you a convoluted (and undocumented!) workaround for getting the original behaviour of the attribute back.

I'm not sure which horse these FAANG kids who excel at programming challenges rode in on, but this attitude is RIFE in their product SDK's and API's.

Dijkstra must be spinning in his grave.

All that based on an increase in page submissions. Is that the goal of auto-complete?

The research behind the number (25%) is highly dubious, because the video cited as the source doesn't tell how it was done. At all. Ironically, it follows a section on how MDN is documenting the standards, and how that "contribute(s) a lot".

There are more than 20 different valid field name tokens in the spec, not just “on” and “off”[1]. If web developers actually used the other values correctly then user agents wouldn’t need to use heuristics to figure out the correct data to autofill.

[1] https://html.spec.whatwg.org/multipage/form-control-infrastr...

I'm reminded of an earlier comment: https://news.ycombinator.com/item?id=21083277

> somewhere along the journey of the web autocomplete=off become a default for many form fields, without any real thought being given as to whether or not that was good for users

And here I was, all along, thinking that website authors were in control of how their websites behaved. How silly of me!

It's complicated. Sometimes website authors do silly things that negatively impact users; browsers ought to help (without breaking things!) where possible.

Some daily annoyances that I wish browsers would actively mitigate, in no particular order: js-based redirects, blocking copy-paste, disabling text selection, hijacking the forward slash to open the website's own search function (I'm looking at you, Github), hijacking any of my other keyboard shortcuts that I rely on, creatively breaking my back and forward buttons (yes I realize there are legitimate reasons for some webapps), and overriding the built in right-click context menu.

The things is that there are valid reasons for all those features: JS-based redirects are useful for many webapps, blocking text selection can be useful for some buttons and such, many people do use custom app-specific key mappings, overriding right-click is useful for many apps (e.g. Google docs), etc. etc.

Of course, all those features can also be abused, but that doesn't mean it's a good idea to try and second-guess website authors; that just makes things much much more complicated for everyone.

Any feature can be abused in any platform, and as the web has moved towards an "application platform" – instead of just a document viewing platform – there are many more features with potential for abuse, but it also becomes much more important that behaviour is consistent and predictable.

I'm most definitely not on the "the specification is holy"-side of things, but from what I read in this issue the Chrome team seems surprising tone-deaf to concerns about Chrome's behaviour here. This is general patterns I see with Chrome, which confuses "it works for the majority of cases/users" with "it works well". These are simply not the same things.

Google is allergic to user interaction. No support, no configuration, no choices, no questions, no customization, no power users.


This should be a simple per-site toggle. And/or a per form/input button to do the autocomplete. It usually only matters for the password field, anyway.

In Firefox, shift+right-click brings up the browser context menu.

Arrgh. I wish shortcuts like this were more discoverable. Shift+right click seems useful and I'll probably end up using it sometime in the future, but I would have never known about it if it were not for this comment. Same as ctrl-L, and for 99% of people F5 and alt+left/right. There should be a page listed in the Firefox menu with a list of useful shortcuts, from most useful (F5, alt+left/right) at the top (so that normal people can get some efficiency gains easily) and going to to more obscure ones like shift+right click so that power users like us can... well, be power users.

They are in control of their servers. They are not in control of how the user agent (clue is in the name) interprets the javascript and HTML sent to it.

Firefox used to ignore autocomplete="off" until relatively recently. I'm sick of sites thinking they know better than me -- like those who don't let me paste in the password field. While I understand the good use cases for this attribute, given the binary choice I would rather have it ignored.

That is absolutely stuff that the user needs to be able to override. In fact, anything related to cutting and pasting is not something I think the browser should mess with.

Also: websites that generate a custom login name or password for me so I won't remember it, and then refuse to allow autofill or pasting. Fortunately dev tools are standard on the desktop these days.

(I would really, really like access to dev tools on mobile, though. It's simply necessary to get around some broken websites.)

In best cases there’d be an option to honor the attribute or not.

The weird thing is that there's other teams within Google who offer autocomplete libraries that simply don't work because Chrome overlays it's own autocomplete on top.

The maps team seems to have given up on trying to resolve that.

Chrome team have made a judgement that autocomplete is required and no-one - not even other teams within Google are allowed to override that functionality.

It's weird.

Google really is the new Microsoft

In spirit maybe. But in practice it seems worse.

> not even other teams within Google are allowed to override that functionality

Isn't that how things should work? If other teams within Google had a special way to override autocomplete wouldn't that be worse?

(Disclosure: I work at Google)

My point is that this decision from the Chrome team even breaks Google's own software.

This emphasises the craziness of the unilateral decision by the Chrome team which is essentially saying: "we, the Chrome team are correct and everyone else can go jump in the lake".

I'm not suggesting that Google should have secret special ways to do things.

Angular Material's Autocomplete component is an example of this. Our users would have found that really helpful... if Chrome's autofill behavior didn't break it.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact