[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Approach to atompub implementation for a forum
On 12/12/07 6:26 PM, "A. Pagaltzis" <pagaltzis@xxxxxx> wrote:
> * Eric Scheid <eric.scheid@xxxxxxxxxxxxxxx> [2007-12-12 07:35]:
>> each entry containing a link to a service document for that
>> topic.
>
> Operating this correctly requires a client that expects to see
> service doc links in entries, and to understand what these
> collections mean.
Does the client know what to do if handed a resource of a particular content
type? My web browser hands any application/xml+atom resources to my feed
reader, and my feed reader offers to subscribe any such resources. I can
also hand text/html resources to my feed reader and it simply displays the
page. I would hope that if I hand it a workspace resource it would offer to
the sensible thing (eg. pass it across to MarsEdit?)
All my feed reader needs, at a minimum, then is the facility of displaying
links of an entry in some generic manner. It does :-)
It also handily notices that some links are to vCal or vCard resources and
will instead special treat them by making icon links in the entry status bar
(alongside the enclosure link icons).
Last I checked though it's not smart enough to realise that clicking on <a
href="feed" type="application/xml+atom">Subscribe!</a> in entry content
shouldn't be handed off to my default web browser (which then immediately
flips it back to my feed reader which then offers to subscribe). This is
probably because the concept of navigating a web of feeds via links in feeds
hasn't yet caught on much .. there's still the assumption that links in
feeds would/should lead to web pages.
> Oh yeah, and all schemes that involve collection-per-thread imply
> that following N threads requires polling N collections. In
> contrast, if you just rely on RFC 4685 for threading, you can
> stick as much or as little data as you want into any particular
> feed. The most natural model is one feed per forum; this puts a
> reasonable limit on the likely number of subscriptions per client
> while at the same time limiting the amount of data in any one
> feed to a likely reasonable volume. So following the most obvious
> approach gives you something that works well.
There's one set of forums I follow where there are about a hundred forums,
and each forum has about a hundred new articles posted daily, and some of
those articles have hundreds of comments posted (and some a relatively few
comments but are nonetheless particularly interesting). Do the math and
explain why I want to download all the articles and thousands of comments
for one forum when I'm only interested in some of the articles. Another nice
effect of subscribing to particular articles is that it's also a form of
bookmarking that article.
e.