[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: non-solutions for blogs.msdn.com: (was RSS consumes too much bandwidth)
Mark Nottingham said:
> I'm really excited to see people advocating delta encoding, and
> advocate it being highlighted as a mechanism of interest to Atom
I don't see why some diff-like scheme is necessary (is this what delta
encoding is?) unless it's necessary to reduce the traffic generated by
within-entry changes. Surely it's sufficient to not send entries to
aggregators that have seen them before? (I'm assuming that most
aggregators these days cache previous entries.)
And I think it should be possible to achieve this with nothing more than
the if-modified-since header: if a client sends an if-modified-since of
"15 minutes ago", then the server should send them a feed containing
(most of the time) no entries. (Or a 302.) If they're lucky they'll get
one. It's not necessary to send them all the entries published more than
15 minutes ago as well; the aggregator has already seen them.
I appreciate Dare's point about it being not a good idea to require
servers to do the equivalent of an SQL query for every request, but don't
think it's necessary for the server to work quite this hard.
If, for each blog it publishes, the server produced and cached "feeds"
that consisted of all blog entries (say) less than an hour old, less that
eight hours old, less than a day old, and less than a week old, then a
client that claimed an if-modified-since of 15 minutes could then be sent
the one-hour-old feed; clients sending no if-modified-since header would
get the full (week old) feed.
(Servers could encourage clients to support if-modified-since by, say,
only including entries more than two hours old in the "weekly" feed. I
guess they could also serve it slower, too, though that's probably more
difficult to arrange.)