Re: [blogite] Before we update the spec...

Date view Thread view Subject view Author view Attachment view

From: Simon Willison (
Date: Sat Sep 07 2002 - 12:25:09 BST

At 13:14 07/09/2002 +0200, mort wrote:
>On Sat, 2002-09-07 at 12:46, Simon Willison wrote:
> What do
> > people think of the system I specced out here?
> >
> >
>I think it's a great framework!
>I made a couple of questions earlier but they went unanswered

My apologies - I meant to answer that mail but Eudora froze and by the time
I reloaded it I had forgotten what I was doing :/

>so if my blogware doesn't have a local copy of this server repository,
>downloads it from a sort of canonical source. Then, either
>a) the canonical(s) server(s) ping my blogware when any record is
>inserted / deleted / updated. My blogware can then choose to sync itself
>incrementally or not.
>b) my blogware downloads a fresh copy of the repository at certain
>times (every hour / day / week / month) to be sure it's up to date.

I think it will have to be a client-pull rather than a server-push thing.
If the server had to ping every blog on the list whenever a new blog was
added /to/ the list things could get messy. Clients downloading the full
list once a week would seem to be a more sensible option.

There's also the third option - the central server has a method to return
all PingBack servers that are "interested" in a URL. Blog clients can
therefore send a list of URLs in a post and get back a list of servers that
they should ping (which they can add to their local copy so that next time
they "see" those URLs they won't have to make a request).

>Other than that, i also think that auto-discovery systems could log the
>servers they find and transmit those logs to the repository. So the
>repository can discard the already known ones and add those really "new"
>This way the repository works for the blogging systems and the blogging
>systems also work for the common wealth :)

Good idea. <link> auto discovery could also theoretically be handled by a
spider / crawler operating on behalf of the central repository. The problem
here is that the central server methods relies on knowing the "URL
patterns" that are accepted by each pingback server - for example, my
server accepts any URL of the format:[*]

In its current format, the <link> element does not provide this
information. I can think of two methods of fixing this, but both require
changes to the spec (which is not much of a problem at this early stage but
could become one later on). Firstly, pingback servers could implement a
second method:

pingback.getURLPatterns() - returns an array of URL patterns that this
server will accept

The crawler could then call this method on any <link> indicated PingBack
servers it finds to get a list. The other option is to change the <link>
element so that instead of pointing directly to the server it points to a
descriptive XML file instead:

<link rel="pingback" type="text/xml" href="/pingback.xml" />



    <server host="" path="/~cs1spw/pimgback.php" port="80">


This matches the format described in my earlier mail.



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 - 13:05:00 BST