Software Architecture: The Hard Parts - Neal Ford

Architects often look harried and worried because they have no clean, easy decisions: everything is an awful tradeoff. Architecture has lots of difficult problems, which this talk highlights by investigating what makes architecture so hard. At the of core many architectural problems: getting good granularity, which we illustrate via event-driven architectures, teams, components, architectural quantum, and a host of other examples. We also illustrate reuse at the application, department, and enterprise level, and why reuse seems simple but isn’t. We also discuss difficult decisions, how to do tradeoff analysis, tools like MECE lists, and how to decouple services to achieve proper granularity. Architecture is full of hard parts; by tracing the common reasons and applying lessons more universally, we can make it softer.

[flashflexpro]

The most practical video worth a millinon and online for free

[ericlegoubin]

Very interesting conference.

Some of my highlights include:

* Every architectural decision is based on a tradeoff.
* If we haven't found the underlying tradeoff, it's because we haven't analyzed enough.
* The tradeoff has a business translation, so the architecture decision can be driven by the business.

The book is now on my reading list 🙂

[self.flavio]

amazing video! great starter to go further on the related book :)

[AlexAD]

Please fire the cameraman and use AI tracking next.

[mfreeman]

I highly disagree with the speakers assessment that LLMs can't be used to do trade-off analysis in software architecture, you just need to guide the conversation and provide enough data/context.

[foxsnce]

you were prompting it wrong. if you then explain what's going on in your system, it will tell you more and how to make a decision. drives me nuts when people just throw a single prompt into chatgpt, take the response and then act like that's all you get. LOL completely ridiculous. But okay, the premise is software architecture not asking chatgpt for help. Lets go.