Entry deletion and distributed systems

[Danny Ayers <danny.ayers@xxxxxxxxx>]

If someone posts an entry and then later retracts it a local system
can handle this ok, but distributing the retraction may not be quite
so straightforward. Maybe it is feasible to distribute the entry
identifier tagged as 'withdrawn' or even 'expired', but I suspect it
may not be sensible to make anything of that at a global level, as it
could never be 100% reliable and the inclusion of such a system in the
specs may lead to unrealistic expectations.

In other words, I'd suggest that rather than having the logic: 

Post => EntryExists
Post & PostRetraction => !EntryExists

We use:

Post => EntryExists
Post & PostRetraction => EntryExists & PostRetraction (and that's all!)

- it's up to the end-user application to deal with that as it sees
fit. So generally (web-wide) entries will only expire by get old and
slipping into obscurity, not through any explicit (un)assertion.

I believe this is consistent with the monotonicity and open world
assumptions the logicians favour for reasoning on the web.

Whatever, it's maybe something worth bearing in mind during
discussions of caching and passing big data aggregations around.

Cheers,
Danny.
-- 

http://dannyayers.com