I want to clarify why this project exists (as many seem to point out that other projects or methods exist for doing this).
TL;DR; If you think of localtunnel as just a shitty ngrok (or name your project here), you are missing the point and probably don't have the same use cases I do.
1. It was made overnight at some hackathon because I was not satisfied with the other tunneling options I found. They required either an account or some stupid ssh setup. I got to thinking of ways to create a tunnel that simply had an CLI tool and instantly get a tunnel no setup. It worked, I kept it.
2. It is written as a library first, CLI tool second. This means it can be used to create tunnels in a test suite if you want to use services like saucelabs to run browser tests (see https://github.com/defunctzombie/zuul). This is leveraged by projects like socket.io and engine.io (among others). This is perhaps the main reason I keep it around despite there being alternative CLI tools.
3. Both the client and server code are availably and easy to install and use. Companies do this when they want to run their own tunnels for privacy (or whatever their reasons... I don't care).
4. Yes, I know the name is identical to the old ruby?python? one. Whatever. That one seems defunct now anyway.
ngrok doesn't have a programmatic API, but I'd love to add one soon. I've built out a library for this in https://github.com/inconshreveable/go-tunnel that will be the foundation for ngrok's next version providing a library in addition to the CLI tool.
Unfortunately, one of Go's weaknesses is that it doesn't embed into other languages like C, so I'd need a ground-up rewrite (in C, probably) with bindings to other languages.
If ngrok the command-line tool had a well defined programmatic interface (like RESTful JSON) would that useful, or is the burden of a separate binary/process to manage still too painful?
To me the ease of having the library be installable with the "canonical" package manager of my platform is too convenient; just "feels" more natural and simpler. I actually thought about writing a node.js ngrok client but then gave up on the idea since localtunnel was working well enough and I personally didn't need the other features from ngrok which I didn't have in localtunnel.
I wouldn't worry about the whole rewrite in c thing. If your server protocol is simple enough, writing clients in the native languages will be better than writing bindings. Installing bindings trips up a lot of users that are not used to compiled software.
Make sense and I agree. Thanks for the feedback! Unfortunately, ngrok's new protocol is optimized very heavily for speed which comes with a cost in complexity of both the protocol and the clients that implement it.
I've never heard of ngrok, but the instantly obvious use-case is to allow testing of webhooks to my local machine. In the past we've done this by booting a temporary server on AWS and remote forwarding to our local machines, which is quite a bit more complicated. I already work on a node stack, so the npm install is wonderful. I expect I'll get a lot of use out of this. Thanks!
I'm confused, there is an existing project called localtunnel that does exactly the same thing and dominates search results for "localtunnel". At the very least, pick a different name.
This seems like a bad idea. localtunnel.me is redirecting non-tunnel'd subdomains to its main page, while inactive tunnel'd subdomains return "localtunnel error: no active client for 'adbc'". So, with a little poking, you find that tunnel'd subdomains seem to be [a-z0-9]{4}.localtunnel.me ... which isn't too terribly large of a search space to crawl. If it gets popular, it should be easy to find works-in-progress that might give up access to the user's computer, or keys to prod, or any of the other stuff that people are a little sloppy about on their work machines.
edit: I was wrong, I should've been a little more thorough. Looks like it's [a-z0-9]{4,10}.localtunnel.me, which is significantly larger.
I've already seen PHPMyAdmin login and tokens pass by, so I don't think it's only used for showing progress but also to have a reverse tunnel back home..
Once you let the public use it, there are certain expectations. That's like saying that an unfinished nightclub doesn't need fire safety because it's a work in progress. That's not acceptable if you're still letting people in to dance.
Pagekite dependency on Python was deal-breaker for me a few years ago. If they would have clients in multiple programming languages so it can be easily embeddable, they would get a lot of business. Prices could be easily higher too.
Also: Vagrant has added a "vagrant share" command that publishes access to your vagrant box, which should be safer than publishing access to your full machine.
Personally I would rather just use "ssh -R", the built in remote port forwarding. You either need to flip a setting on the server to allow listening on an interface besides localhost, or configure Nginx/etc as a reverse proxy.
That's what I've used in the past, but its more complicated to set up ("You either need to flip a setting on the server to allow listening on an interface besides localhost, or configure Nginx/etc as a reverse proxy." is not trivial for most). It also requires a server with a static IP to ssh to.
Yes. This is cold hearted, but I would suggest that for those for whom this is too hard, they should reconsider exposing their machines and code directly to the public Internet.
Why should they reconsider it? Someone that just got into development and might want to show their friends or family something they are hacking away on. According to you they now need to learn about VPS, some random nginx settings or other SSH nonsense and meet some arbitrary "minimal" criteria you have decided upon because that is how you would do it. I think you should reconsider your acceptance of people that don't share your same technological expertise.
localtunnel v1 worked perfectly for me, localtunnelv2 never did (apparently no remote server was ever up) but it's weird that ngrok is a ruby project again.
This seems like a bad idea or to phrase it correctly: use it wisely.
Because, you will use this service for the development to give someone outside access to something. If you then close the tunnel, the service will forward any request to its own server either to the main pager or to an error page. That means, all data given with a request, either via GET or via POST, will be given to that service. That could include sensitive data. That means, this kind of service is security risk.
There's also http://httpi.pe/ - pretty similar both in concept and in implementation. The major difference would be an 'inspection' view allowing users to view the traffic going through the tunnel.
Or at least get to reuse an URL after reboot/dc/something, since this is links you would send to your client/boss, having to email them new ones is a hassle.
I want to clarify why this project exists (as many seem to point out that other projects or methods exist for doing this).
TL;DR; If you think of localtunnel as just a shitty ngrok (or name your project here), you are missing the point and probably don't have the same use cases I do.
1. It was made overnight at some hackathon because I was not satisfied with the other tunneling options I found. They required either an account or some stupid ssh setup. I got to thinking of ways to create a tunnel that simply had an CLI tool and instantly get a tunnel no setup. It worked, I kept it.
2. It is written as a library first, CLI tool second. This means it can be used to create tunnels in a test suite if you want to use services like saucelabs to run browser tests (see https://github.com/defunctzombie/zuul). This is leveraged by projects like socket.io and engine.io (among others). This is perhaps the main reason I keep it around despite there being alternative CLI tools.
3. Both the client and server code are availably and easy to install and use. Companies do this when they want to run their own tunnels for privacy (or whatever their reasons... I don't care).
4. Yes, I know the name is identical to the old ruby?python? one. Whatever. That one seems defunct now anyway.