From: Ian Hickson (email@example.com)
Date: Mon Sep 09 2002 - 21:59:56 BST
On Sun, 8 Sep 2002, Jim Dabell wrote:
> > we have already established that it is possible to write XML-RPC
> > proxies for this kind of stuff instead.
> It's possible, yes, but I don't like proxies much more than centralised
> servers. It adds complexity that can't be simply programmed away.
Ditto with an XML layer.
> Which proxy?
Whichever, it just has to be an appropriate one.
> What if it breaks?
What if the server breaks? What if the client breaks? What if your XML
file is malformed because of your ISP slapping HTML code into it?
> What if it silently drops half my notifications?
> What if a bug sends me the wrong information?
This applies to all of the clients, not just proxying clients.
> What if you decide to shut it down?
Run the proxy on your own machine, the source is available.
> What if... all sorts of things that you don't have
> control over because it's not your server and it isn't guaranteed to work
> in any particular way (as opposed to servers that have to follow the spec).
A proxy _is_ a server, so it has just the same guarentees as any server.
(Indeed, my server proxies comply to the spec better than my blog, since
my proxies support sending back the standard fault codes, which my blog
> I see it as merely shifting the programming work from the blogging
> tools to the proxy tools, and incurring the penalty of a third party
> in the process.
But your system increases the work of the blogging tool authors, and
incurs the cost of greater barrier-to-entry, instead.
> I honestly don't see this spec as being that complex, particularly
> because you can implement the bare minimum, and come back to add more
> methods later.
Well, I would not have jumped forwards and implemented the spec if it had
required me to hit a remote server _three times_ (link, XML, server) just
to tell the server "hi, I linked to you".
Ideally, a single hit should have been all it need take. HTTP 1.0 even
supports that, using the "LINK" method (not to be confused with the Link
header, BTW). Or a new method could have been invented. Then the HTTP
transaction would have looked a little like:
PINGBACK /~cs1spw/blog/archive/2002/09/02/#pingBackImplemented HTTP/1.1
No need for anything even as heavy duty as XML-RPC.
Unfortunately, doing that on static blogs is a little harder than most
people are ready to accept, I imagine. (My blog could implement that in
about 10 lines of code though!)
> Remember, 99% of people won't be programming this stuff, they'll just
> be using off-the-shelf blogging tools.
Maybe, but I'll be implementing it! :-)
> I think that servers automatically talking to each other and
> establishing links is an interesting concept, but it breaks when you
> have to hard code a third party in to do the talking if you want
> something slightly different from normal.
Well, the question about why you would want something else remains. The
only case I can see is that you can't put scripts on your domain, so you'd
want to use e-mail instead. But we've already knocked down the idea of
automatic e-mails, apparently, so what else is left? If you can run
scripts, then you can either plonk one of my proxies on there or write
your own XML-RPC code.
-- Ian Hickson )\._.,--....,'``. fL "meow" /, _.. \ _\ ;`._ ,. http://index.hixie.ch/ `._.-(,_..'--(,_..'`-.;.' Message sent over the Blogite mailing list. Archives: http://www.aquarionics.com/misc/archives/blogite/ Instructions: http://www.aquarionics.com/misc/blogite/
This archive was generated by hypermail 2.1.5 : Mon Sep 09 2002 - 22:05:00 BST