[wp-hackers] Forum Help

[Matthew Thomas]

Matthew Mullenweg wrote:
> Matthew Thomas wrote:
>...
>> Using POST buttons where appropriate, instead of links, would fix all 
>> those problems, but I guess there's some good reason for not using them.
> 
> Finally! Thanks for understanding that the real world may appear simple 
> from the heights of the ivory tower but actually there's usually a good 
> reason for everything in a mature system even if it's not immediately 
> apparent.

Dude, that was the opening for you to say what the reason *was*. :-) If 
it was ever discussed, sorry I missed it. I seem to recall suggesting 
pre-1.2 that "Log Out" should be a button (though the "Delete" links 
escaped me completely), but I can't remember why you disagreed.

This is something that's been discussed in the What-WG recently
<http://listserver.dreamhost.com/pipermail/whatwg-whatwg.org/2005-May/thread.html#3897> 
-- why are people using links for non-idempotent actions when they 
should be using POST? The prevailing theory is that it's to make them 
look pretty compared with surrounding links; but that can't be the case 
for WordPress, because the "Delete" links are in table cells by 
themselves, and other buttons in the admin interface are heavily styled. 
So seriously, I'd love to know.

> Google, for all their strengths, was idiotic for releasing a product 
> that breaks the way millions of web applications, including some of 
> their own, work. Note that no one from Google came out citing RFCs as 
> the reason their product is wreaking havok across the web, because they 
> realize more than anyone else that GWA is broken, which is *why they 
> took it down it after just a few days*. I find it funny that so many 
> people are rushing to their defense -- they're a big company and can 
> defend themselves.

I'm not defending them -- serving people *other* people's cached 
versions of a page, in a way that existing proxy caches don't, is asking 
for trouble. But the prefetching is an interesting example of how, if an 
allowance in a specification isn't taken advantage of quickly, it tends 
to become unusable. If prefetching accelerators had been this 
well-publicized years ago, authors would have realized the problem 
sooner, and there wouldn't have been so much pain now.

-- 
Matthew Thomas
http://mpt.net.nz/