[rfc-i] Regarding RPC effort in producing XML RFCs

[Martin Thomson <[e-mail]>]

The latest consultation from Jay on RPC throughput highlights a few areas where it seems to me like the RPC is doing work that authors should be doing.  Though it makes sense for the RPC to be responsible for the final quality, that doesn't naturally mean that they need to do all the work necessary to meet that quality bar.  Authors are part of this process.

In particular (quoting):

● Use of id2xml to convert plain text submissions
● Converting artwork into RFCXML tables, sourcecode, lists etc (and sometimes vice versa)
● Converting/correcting lists to use the correct form (definition lists in particular
● Identifying and applying source code types

Would it save the RPC a lot of time if authors were responsible for much of this work?  The RPC could insist that documents they receive are in XML, with native tables rather than ASCII art, annotated sourcecode blocks, and lists in the correct form.

Reviewing is likely to still take time, but far less than doing the work.  Automated checks (hello idnits v3) could flag these things, so that authors have an action plan before the RPC even sees a document.

Now, I could suggest that finding ways to use LLMs to do some of the manual work of the RPC is also worth exploring.  Some of the reviewing process is well-suited to LLM use. A very cheap&fast model could do much of the checking and even fixing.  Still, for this sort of thing, authors can do this work.

The length of the queue doesn't need to be zero, but it does need to shrink rather than grow.  Having the RPC do *less* has to be part of any solution.