I started another specialized thread about this since the last one
("Proposal to add HTML class attributes to SRFIs to aid
machine-parsing") became so broad and detailed that it was getting
hard to navigate.

To summarize: Discussion advanced to the stage where we could either
update the SRFI HTML files with a few new HTML class attributes, or
transform them altogether into a strict XHTML form. Either form would
allow us to auto-extract metadata (general information, symbol index,
procedure argument names) into an S-expression format. Additionally,
advanced type signatures for static type inference could be written by
hand. These could also be in S-expression format, and could go into
either the same or different repos/files as the auto-extracted
metadata (which are planned to go into the SRFI origin repos).

At this point we could start to design the precise syntax for the
S-expression file(s). And together with the syntax, the set of
information that goes into the file(s).

The following general information about each SRFI could be included:

* number: 158
* title: "Generators and Accumulators"
* authors: ("Shiro Kawai" "John Cowan" "Thomas Gilray")
* status: final
* mailing-list: srfi+minus+158+at+srfi+dotschemers+dot+org
* mailing-list-archive: https://srfi-email.schemers.org/srfi-158
* history:
     (2017-08-11 "Received")
     (2017-08-12 "Draft #1 published")
     (2017-10-09 "Draft #2 published")
     (2017-10-21 "Draft #3 published")
     (2017-10-27 "Finalized")
     (2017-10-29 "Fixed description of gmap")
* abstract: "This SRFI defines utility procedures that create..."

We might be able to auto-detect that the license it the MIT license
(has any other license ever been used?)

Additionally, we can build a symbol index with few constraints on the
HTML markup.

Comments? Ciprian, had you already worked on an S-expression skeleton
with the above general information included?