Re: ERROR verb

[Asbjørn Ulsberg <asbjorn@xxxxxxxxxxxxxx>]

On Sat, 19 Jun 2004 15:34:15 +0100, Bill de hÓra <bill@xxxxxxxxxx> wrote:

> Just that ERR is the verb form of ERROR.

Ah, of course. *Smack forehead*

> I responded to Henry about this - using exceptions for control flow is a  
> questionable technique.  I don't think it's RESTful - it's relying on  
> assumptions about (invisible) behaviour and the nature of  
> implementations, such as log file analysis, or even that servers have  
> log files.

Well, it's not that important anyhow. I mean, if you ERR an URI, and the  
web server doesn't log the request, no one knows about the ERR. No harm  
done. What Atom, and the rest of the XML community would do, is spread the  
word about ERR and try to get it into an RFC as an amendment to HTTP. As  
Hentry wrote, this problem isn't only with Atom, but with all XML on the  
web. All.

We need some kind of standard error reporting mechanism. XHTML served as  
'application/xhtml+xml' needs it maybe even more than Atom. I think that  
if we go together, define a good spec for this (it would be crazy if it  
were on more than 2-3 pages total) and then spread the news about it,  
Apache and other web servers would have support for it in not so long  
time. And in the meantime of no support, ERR request will at best be  
logged or handled by custom HTTP handlers, and at worst not handled at all.

I think the solution is perfect. It already works now, it's easy to  
implement, it doesn't break anything, and it serves the whole web  
community, not only Atom.

> Adding any new verb is a huge requirement.

If we add this to HTTP, it won't be Atom's requirement ;). And if

> And that Atom is capable of doing more has yet to be demonstrated
> (to me ;).

I love the format inside-out, but I almost agree with you. The  
specification already gives probably ten times more than both of the RSS  
specifications together, but the sole formats alone don't give much from  
each other. They all do practically the same thing. That's why I think  
it's important that Atom has e.g. threading support in v 1.0, so that it  
actually has something to brag about.

> If the intent of the method is that a client report an error to the  
> server, then you will be making implementation assumptions beyond HTTP.

True. Therefore, let's put the verb on top of the HTTP stack of  
specifications instead. Seriously.

--
Asbjørn Ulsberg         -=|=-        asbjornu@xxxxxxxxxxx
«He's a loathsome offensive brute, yet I can't look away»