So one of the issues with Git is that just about any Git related command ends up being very chatty in terms of disk IO, i.e. a git show whatever isn’t just a single IO operation, it’s a whole bunch of them. When you end up running a forge for many users, this can wreak havoc in terms of performance and outages. Loading non-Git data (e.g. issues) from Git makes this even worse, especially if a project the size of say Kubernetes were to use it (e.g. they have about 50 000 issues in total).
How would Haxy handle that? That’s not some sort of “haha gotcha!” question, rather I’m seeing new forges pop up over the last two years or so but few of them seem to actually try to plan ahead/learn from all the scaling horror stories of GitHub, GitLab and others, so I’m curious if Haxy is any different in that regard.
score 1 · canonical
A big part of it is the complexity of the specification. The various HTTP 1.1 specifications are by no means trivial as there are a lot of small details developers are likely to forget about, but HTTP 2 is a lot more complex, such as due to it’s header compression algorithm amongst other things.
As for HTTP 3: while HTTP 2 does provide clear benefits over HTTP 1.1 for end users/efficiency, the difference between HTTP 2 and HTTP 3 is a lot smaller and honestly just not worth the trouble, unless maybe you’re operating at Netflix/Cloudflare/etc scale
score 1 · canonical