Hello!

A while ago there was discussion about two link extensions I-D:s by James Snell:

* Link Extensions
<http://tools.ietf.org/html/draft-snell-atompub-link-extensions-02>,
and
* Tombstones <http://tools.ietf.org/html/draft-snell-atompub-tombstones-05>.

I'm quite eager for these to become RFC:s. The project I'm working in
is getting close to roll out a big data integration project between
swedish government agencies. Currently I use extensions from these
I-D:s to manage a flow of entry updates and deletions (both of whom
will be rare but necessary).

Primarily we need the tombstones. Additionally, we use the "md5"
attribute (and possibly the "etag") as a compact way of providing an
explicit means of checking the correctness of transported documents.

We use this very instrumentally, and we could define our own
extensions of course. There are also more or less usable alternatives
in the wild, but (from what I've seen) those fit the bill much less
precisely.


* For Tombstones, just using GData:s <gd:deleted/> marker in entries
does do the job. So does the "deleted" attribute on the sync element
defined in FeedSync (AFAIK). But these other specs incorporate *much*
more than what we need, and I'd prefer not to overload the respective
feed producers with the burden of filtering those specs just to get at
this very specific need. (Our working implementation handles all three
kinds; but I'd prefer to remove this choice.)


* Regarding the Link Extensions, as mentioned we only use one (or two)
of them. But I haven't found much of comparable simplicity and with
the quality of specification I (IMHO) perceive this I-D has. (Though
I'd perhaps prefer if it was possible to supply checksum and algorithm
separately, as seen in some of the alternatives mentioned below).

There's the Yahoo Media RSS extensions, or going all the way to embed
XML-DSIG-chunks just to mark the links with checksums (I think, the
semantics of DSIG:s seem very complex for this simple case). I also
recall seeing some extensions for md5/checksum while "googling".
Still, neither have seemed better to supply to the feed producers.

(Admittedly, if GData had also defined a checksum extension (for both
atom:content and atom:link) I'd probably just go with GData pieces for
all of this, for better or worse.)


One resort might be to keep on using the extensions of these I-D:s
regardless of that such a practise is formally inappropriate. That or
just remint namespaces for them and keep this usage "in-house" (as big
as this "house" might be). However, this will divert what would
otherwise be (what I hope to be) a concrete usage, which could promote
wider adoption if these where already accepted RFC:s.

(I also fear that an effort from us (we have limited resources) to
specify equivalents of these ourselves and publish publicly is much
less likely to be adopted than these already ongoing I-D:s.)


Is there any way forward in this? I'd love to supply time to get these
accepted as RFC:s, but I don't know (at all) what that would entail.
Or is there simply not enough interest to push them forward?

Best regards,
Niklas Lindström

Reply via email to