RFC for the layered RPC system

[SirVer]

Rendered version

[SirVer]

@lastorset ping - as requested on my blog. Very interested in your comments.

[gchp]

Here's the link to the rendered version A little easier to read.

[mmatyas]

I have some questions:

• Is 16 bits really necessary for priority?
• In case some plugins have the same priority, how are they sorted?

[mkpankov]

@mmatyas

My thoughts on 16 bits for priority: I think it makes sense. In edge case you'd like each plugin to have distinct priority. And, say, 8 bits is not so much in this case: I currently have Emacs setup with nearly a hundred packages, so I can imagine somebody has > 255 of them. 16 bits is much more future-proof IMO, especially in case there would be an actual Swiboe server running remotely and servicing multiple users (not sure if that's in current design though).

[SirVer]

@mmatyas Those are very good questions. I posted some clarifying thoughts in 6db81db. Mostly, my thoughts mirror @mkpankov's.

The idea of priorities came from UltiSnips. Before priorities, the order of when your snippets were loaded defined which snippet overwrote which other. This was problematic, because of lazy-loading of plugins, reversed orders of plugin path traversal for pathogen and vundle. now every snippet has a priority and order is clear.

[mmatyas]

Ok, thanks for the explanation! What I'm afraid of is that many plugin creators may simply leave priority on default, and if that may cause problems. I wonder if you've got similar issues during working on UltiSnips?

[SirVer]

I think leaving priorities at 0 is fine for most plugins anyways, so I am not worried about that. It is important that there is something that can be tweaked if there are collisions.

In UltiSnips I had one or two bug reports of people not getting the priority system, but none where it made problems. I think for the RPCs, priorities are a bit more intuitive than for snippets, so I think that will work out too.

[SirVer]

Seems like there are no more comments on this RFC. I'll leave it open for now and will merge that in on the weekend if there are no more comments.

[SirVer]

@homu r+

[homu]

📌 Commit 5e6477c has been approved by SirVer

[homu]

⌛ Testing commit 5e6477c with merge fc308e0...

[homu]

☀️ Test successful - status