Tim Bray wrote:
> 1. Should we use URI templates? (the current draft only has {index},
> and it seems unlikely that we could get consensus on anything more
> ambitious).
-0.5
I really think this should be an extension. I'd be OK if they're part of
the core but only if optional.
> 2. Should we have prev/next pointers in collections to allow clients
> to page through?
+1
> On the face of it, it seems like all four possible outcomes are
> plausible.
> - No templates, no paging
How would it work?
Even protocol-04 and gregorio-09 had paging!
> - Templates but no paging
-1
Even with templates, servers might want to limit response length and then
provide paging on the result "feed" (e.g. client wants entries 1-10000,
server sends only a 10-entries long feed document with "next" link to the
next 10 entries and so on until the client can get the 10000 entries)
> - No templates but paging
+1
This wasn't a problem in gregorio-09 and protocol-04, until James started
talking about search functionnality and people started to limit search to
only filtering capability, then shrink filtering to index or date ranges
only.
> - Both templates and paging
-0
Only if templates are optional. But if they are optional, there's no need
to have them in the core: just put them in a "companion RFC" (similar to
what RFC2617 is to RFC2616).
I'm not trying to "just remove" templates. I think RFC3229 w/ feed is more
appropriate for syncing and the paging is enough for the core. However
templates are a good thing, simpler than RFC3229 w/ feed, so they should
be made "official" at the time as the core or a few weeks/months after.
--
Thomas Broyer
