Re: [blogite] Re: b2 client pingback

Date view Thread view Subject view Author view Attachment view

From: Simon Willison (
Date: Sat Sep 07 2002 - 16:28:22 BST

At 15:55 07/09/2002 +0000, Jim Dabell wrote:
>Is an entry with 20 links really that common? Even if it is, surely it
>would be a much better approach to do this processing *after* you have
>returned the page to the client (i.e. all you do on the submitting page is
>schedule an at job for a couple of minutes into the future, or put an entry
>into the database for a cron job to notice).

This is more of a problem with PHP than anything else. As far as I know,
the only way to make PHP return information to the client before it has
finished processing is to use output buffering and "flush" output straight
to the client while the script is still going. This is great, except I'm
pretty sure (not certain mind) that if the user terminates the connection
by hitting "stop" the PHP script will also terminate, which would kill off
your pinging process mid flow.

> > With a
> > locally cached list of blogs that are interested in specific URLs that
> > overhead is gone - all I have to do is ping the few pingback servers (if
> > any) that my list tells me want pinging. A central server is therefore a
> > useful addition (not a requirement mind) to the system as it acts as a
> > place I can get a list of blogs from at regular intervals.
>It's also a failure point. What if the cache is outdated, for example?

Good point. The idea is that blogs send a ping to the server asking to be
added to the list. This is great, except it means that for the central
repository to be of any good it has to be a requirement of pingback enabled
blogs that they register with it - and we don't want to add anything like
that as a requirement of the system.

It seems it's just myself and mort that are interested in the centralised
service idea for the moment, and the arguments against it are pretty good.
That combined with the realisation that the overhead isn't quite as bad as
I initially thought (you only have to retrieve pages up to the </head> tag)
has convinced me to stick the central server on the back burner for the
while and concentrate my efforts towards sorting out the stuff everyone
agrees on (the basic spec and <link> auto detection).

If it turns out a centralised service is required we can set one up later -
I've got the code for it already.



Web Developer,
Message sent over the Blogite mailing list.

Date view Thread view Subject view Author view Attachment view

This archive was generated by hypermail 2.1.5 : Sat Sep 07 2002 - 17:05:01 BST