Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: Are there any openly available software architecture documents?
407 points by passenger on Jan 10, 2020 | hide | past | favorite | 49 comments
I'm about to leave my job for another and i'm tasked with documenting the existing architecture and flow of the app.

I'd also have to document why certain decisions were taken. I'm looking for any openly available similar documents for guidance mostly around how to structure the document.

There are bunch of them, really.

But for general structure, using your end-user's user experience path (from start to end) as a guide and avoiding buzzwords as much as you can usually help. Think of Stripe's documentation, you want something as easily digestible as that.

- Software Design Patterns: https://en.wikipedia.org/wiki/Software_design_pattern

- Azure Application Architecture Guide: https://docs.microsoft.com/en-us/azure/architecture/guide/

- Azure Cloud Design Patterns: https://docs.microsoft.com/en-us/azure/architecture/patterns...

- Azure Architecture Framework: https://docs.microsoft.com/en-us/azure/architecture/framewor...

- Azure Cloud Adoption Framework: https://docs.microsoft.com/en-us/azure/cloud-adoption-framew...

- Cloud Computing Patterns: https://www.cloudcomputingpatterns.org/

- Microservice Architecture Patterns: https://microservices.io/patterns/index.html

- Amazon's Builders Library: https://aws.amazon.com/builders-library/

Apache Kafka's documentation is pretty good, too: https://kafka.apache.org/0102/documentation.html

Also, for a 10,000 foot overview of popular open-source tools, see https://aosabook.org

There are dozens at the Architecture of Open Source Applications ebook/site: http://aosabook.org/en/index.html

As someone who only really gets exposure to web tech, it was fascinating how other types of software is architected, too.

The best book/pair of books about software that I have ever read.

Perhaps Effective C++ comes in second, but that's just b/c C++ is a nightmare.

Effective C++ is mostly just a book about C++ pitfalls to avoid. It says nothing about high level concepts.

Wow, what a great resource. Thanks for sharing.

Highly recommend AOSA

Don’t do it like me and overthink it. Just write down all your thoughts informally. Make diagrams by hand. Only once you have most content then think about formatting and structure. A lot of people produce nicely looking docs without much content. I much prefer a bunch of simple notes that describe the thought process

I agree. I used to work on a project where everyone involved in the initial design had left the company, and documentation was spotty. The major issues I had working with the code were all related to not understanding the underlying design choices, e.g. why a certain model was chosen over another or how a certain abstraction was intended to be extended with new functionality.

Stuff that can be formalized in a systematic manner is usually stuff you can figure out on your own in a text editor or IDE.

We've been capturing those decisions in ADRs (https://www.thoughtworks.com/radar/techniques/lightweight-ar...) for the past year or so, and they've been incredibly helpful even for those involved in the original decisions that want to revisit the rationale - I can only imagine how helpful they'd be in a situation like yours.

So much this. Working for Enterprise I saw so many 20 page architectural documents that were borderline impossible to parse and were about as illuminating as a 30 second conversation.

But then it is all .. vague. I do hope something better than can be done.. hand drawn diagrams can't represent details at different levels of magnitude, like for example why statecharts are very expressive compared to other diagrams is it's ability to have arbitrary levels of nesting, which is not possible with handdrawn or even printed diagrams.

I feel like resorting to simple hand drawn diagrams, is in a way declaring failure on being able to describe complex software requirements and structure/ behavior in a systematic graphical manner.

What works for me is to do things by hand, until all the possible information I might want to share is present. I call it my inventory.

Then work on refining, editing, deleting, modifying.

Separate the creation from the editing/communication.

This x1000

Just make sure your stuff really expresses something and isn’t just pretty as I see quite a bit.

The details will always have to be looked at by the inheritors - it's more important to not say something wrong or contradictory IMO. When I've had to do something like this in the past I circle through it, documenting the high-level ops before details anywhere. Tricky or especially complicated details next, then the operational details for as much as I've time for.

I'm doing this now for a couple of things I've inherited because my predecessors either never documented it or it's been lost to time. High-level flow, tricky and/or problematic bits with operational & recovery info, then the mundane as I have time. That means the mundane or some of the less-tricky tricky bits will be "vague" for a while.

You're absolutely right about not overthinking. Any architecture doc is better than none. Even if the thinking it describes is incoherent or if it describes coherent thinking sloppily. And even during editing, there's only one thing I want my principal engineer to keep in mind. Inside the technical vision, tell me a little about implementation details. Either pulled in from below or, if the principal is risen from my ranks, written from experience.

C4 is pretty easy to learn, Arc42 has some sample documents that are very useful. They complement each other.



PRINCE 2 documents also come with a 'decision log', if you want something in a formal template.

The book `Documenting Software Architectures ...` [1] can help.

[2] and [3] list a few example documents for Software Architectures, you can google further and find some more.

I think reading through a few sample documents and based on the expectations at your current job, you can derive the necessary doc.

[1] https://www.amazon.com/Documenting-Software-Architectures-Be...

[2] https://projects.cecs.pdx.edu/attachments/download/3146/Soft...

[3] https://www.voa.va.gov/DocumentView.aspx?DocumentID=188

AWS has nice whitepapers on their software architecture.


I’m a huge fan of code as design [1], so I tend to do my architecture thinking at a high level - the objects and interactions between systems - and then write clear and simple code which expresses my intention. To this end, if the code is clear then it doesn’t need to be documented heavily.

I’m also a big fan of DDD [2] and use his “box and line” drawings all the time. There are loads of resources for DDD but I’ve linked to the book, which is awesome. DDD has great concepts like Context Boundaries and Ubiquitous Language which may simplify the documentation process.

It’s easy to think of architecture too abstractly or in terms of frameworks or other complexifying concepts, but simple is good, and that’s true from top to bottom.

[1] http://www.developerdotstar.com/mag/articles/reeves_design_m...

[2] https://www.goodreads.com/book/show/179133.Domain_Driven_Des...

These slides do a good job at highlighting the important points from the DDD methodology:


I'd also have to document why certain decisions were taken.

Yes please. So much this. With evidence, please, so that the decisions can be revisited in the future and changed when they no longer apply. Every day I open another source file on a system that was built over a decade by a rotating cast of mostly first job software engineers without adequate mentorship and (I wish I was making this up) having their C++ code reviewed by a C programmer who gave them comments such as "just turn all these functions into one-line macros", and someone who appeared to be mainlining late-nineties OOP hype like it was going out of fashion and needed to get it all on the page before someone beat him to death with an old Vic 20.

Every day we stare at it and ask ourselves why. Why was it done like this? Was there a good reason back then? Does that reason still exist? If I rewrite it to remove all this horror, will something else break? Even, oh God, even "What is this actually meant to do?" I can see what it does do. What it does do makes no sense. Did it ever make sense?

Even the occasional comment like "cheeky hack for performance" is often more hindrance than help when a quick test shows it actually makes performance worse. Did it ever improve things? Did the hardware change underneath it such that it no longer helps? Did other code change such that it no longer applies?

Anyway, where this was going is that a decision recorded without evidence or even the factors considered might as well have been a coin toss to the software archaeologist having to maintain it in a decade's time. If the decisions and reasoning aren't recorded, you're forcing someone to attempt to reverse-engineer your state of mind, and that's an expensive and impossible job.

Plantuml for drawing pictures together with the C4 Plantuml Plugin is helpful. i.e. you make high level overviews of your architecture and then „zoom in“

For decisions a table with some explainatory comments why you have decided to use this one thing and which alternatives you have considered and why those did not make it, helps. You can look up ADRs. (Decision Records)

There is the excellent Architecture of Open Source Applications book, which is available online for free at http://aosabook.org/en/index.html

It contains architecture overviews of many open source projects.

Our design decision documents have essentially four sections: Problem description, solution options, comparison of the options, and decision. I find it helpful to clearly distinguish what is part of the problem and what part of the solution. The size of the sections varies a lot.

Meta information could be: Status, stakeholders, decision-makers, and rollout description. It also often has links to meeting protocols and tickets.

These are not architecture documents but change management documents. Roll out of such decisions should, among other things, result in an update to architecture docs.

This might initially look to be be a little too specific for military / aerospace, but it can act as a good guideline for a table of contents and examples of what might go in each section.


The US Military came up with their own standard for systems and software development (MIL-STD-498), which as evolved into IEEE and ISO standards. The nice thing about the US Military's standards is that they came with Data Item Descriptions (DIDs) that specify exactly what goes in each document.

I am a fan of Phillipe Krutchen's[1] "4+1" model[2], or variants thereof. The book Software Systems Architecture: Working With Stakeholders Using Viewpoints and Perspectives[3] by Eoin Woods and Nick Rozanski is a really good overview of the overall mechanism.

Also, while this bit is probably overkill for your current situation, you may find value down the road in looking at some of the various "architecture frameworks"[4] and reference models that are out there. Things like TOGAF[5], DODAF[6], MODAF[7], TRAK[8], etc.

In any case, always remember the words of the immortal Bruce Lee:

“Absorb what is useful, discard what is useless and add what is specifically your own”.

[1]: https://en.wikipedia.org/wiki/Philippe_Kruchten

[2]: https://en.wikipedia.org/wiki/4%2B1_architectural_view_model

[3]: https://www.amazon.com/Software-Systems-Architecture-Stakeho...

[4]: https://en.wikipedia.org/wiki/Enterprise_architecture_framew...

[5]: https://en.wikipedia.org/wiki/The_Open_Group_Architecture_Fr...

[6]: https://en.wikipedia.org/wiki/Department_of_Defense_Architec...

[7]: https://en.wikipedia.org/wiki/MODAF

[8]: https://en.wikipedia.org/wiki/TRAK

I found the 4+1 architecture docs very helpful when working with a recent client. https://en.m.wikipedia.org/wiki/4%2B1_architectural_view_mod.... What I like about it is that it separates the architecture into four views each for a different audience

Most important thing to convey is the conceptual framework used and assumptions made to simplify things. e.g. "We will fix problems only by altering the base image, not the build system." "We will not change to 64-bit build as a part of this change." The sort of thing so that if someone really internalizes the perspective, they will be able to decide all the details on their own. Also a basic picture is always very popular. The details should be in git someway or another, but being able to use the details quickly is the job of the overview. I find it useful to document common change workflows ('To add another type of processor, add the class name to this map in this file, provide a jar implementing that class here, and add a new ID to the enum here.')

I've found this to be a pretty great resource: https://github.com/joelparkerhenderson/architecture_decision...

ADR is made for this


There are multiple variants. Simple or detailed.

ADRs are great, and don't need to be overly formal or use tooling. We've been hand-writing Markdown-based ADRs (https://github.com/adr/madr) checked in with our code in a docs/adr directory.

You don't put every little decision in of course, but most projects have at least 1-2 areas where even the people involved with the decision have a hard time recollecting all the context for the decision.

The downside, at least for OP, is that they're usually something you capture at the time the decision is made - not when you're leaving a project.

PlantUML is useful for diagramming anything from sequences to system architecture to state machines. It's also widely supported in many modern text editors like emacs and VSCode. PlantUML components for AWS are widely available [1]. You may find RealWorldPlantUML a good place to start looking for examples [2].

[1] https://github.com/milo-minderbinder/AWS-PlantUML [2] https://real-world-plantuml.com/

I've used the C4 model to draw an overall system architecture (container level + external systems). https://c4model.com

I think it's worth looking at it like another product that you produce. I like to imagine someone seeing my code for the first time and trying to make sense of it.

Architecture: this is often documented with some kind of diagram. I personally find them to be unhelpful.

Flow: I'm guessing people aren't familiar with your UI. The best way to document would be to record a video, but that might be a little extra. I would make a powerpoint doc with screenshots to show the common flows.

For me, it kinda looks like this:

- Assume expertise in the individual technologies you've used

- If you picked the tech, then write why you did, and link to resources to learn more about it

- Write down which general constraints I placed on myself that might not be common. EX: naming patterns, cyclomatic complexity, using a functional approach, or being more point-free in a language where it's not common, choices around duplicate code

- Make a nested list of the folder structure and describe what each folder is for (even if it feels obvious), and describe how the different parts interact with each other. The questions to answer are: what are the high-level dependencies between dirs? Are there any cyclical dependencies?

- Find the knots (especially complex parts of the codebase) and make sure they're documented properly

I agonized over the best way to do this stuff for a while also and the best framework I've used so far is the C4 model (as mentioned elsewhere).

For documenting decisions on architecture this should be done as you go really but better late than never :). Using adr ( architecture decision records ) text / md files would be my approach. They will typically be kept alongside the code and source controlled

A little known, but powerful notation for description and documentation of architecture of Information Processing Systems is FMC http://www.fmc-modeling.org/ This site features a link to the Apache Modeling Project where a now outdated version of httpd was documented using this methodology.

I would also look at "The Architecture Of Open Source Applications", although it might be more detail than you're looking for.

Archi software to draw Archimate diagrams.

Making use of colors in architectural diagrams really add depth to the information. Often applications involve internal and external (out of scope) components. Using common colors for each group of components really helps in the readability.

If you only have a few weeks then there might not be any good results: you have to spend time learning the framework and also provide the relevant/ required info.

It is an opportunity to learn and apply it forward to your new job.

GitLab has all of their docs online..I guess you can dig into them for inspiration

Do you have a specific link one can start digging through?

Tell them the code is so good it documents itself.

import: English

import: Diagram

ok: Community

see: https://www.drupal.org/project/project_module?f%5B2%5D=im_vi...

Yikes, if you haven't been doing this as you go you've severely messed it up.

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