RE: Section 14.36 Range, and PUTs

>----------
>From: 	jg@w3.org[SMTP:jg@w3.org]
>Subject: 	Re: Section 14.36 Range, and PUTs 
>
>Maybe I'm missing something, but wouldn't a PUT + Range be exactly
>symmetric to GET?  The range information would specify which range to
>replace (If multiple ranges, you'd have to transmit a Multipart message
>to contain the ranges).
>
>This would be very useful for updating parts of large objects.

I agree. It also lets you append cheaply.
>
>I don't have much against forbidding it given we've not thought it
>through
>significantly, for 1.1, but wonder if we should be that draconian.

I think its semantics are quite clear, and someone will want them, so I
don't see any reason to forbid it.
>
>This does beg the question of what DELETE + Range should do; probably
>best to say it is illegal.
>
>Any other opinions?

I think its also clear what DELETE+Range should _do_ if it's
implemented.  I don't think its reasonable to require that anyone
implement DELETE+Range, although I have worked on a system that had file
system support for that functionality (it's kind of b-tree like, with
byte offsets as keys...).

Now, POST+Range is what I don't find clear.

However, I don't much care if we make Range illegal (for now) with
anything except GET, HEAD, and PUT.

I'll reiterate my opinion that 1.1 servers that don't actually do
PUT+Range MUST return 501 (Not Implemented) rather than ignoring it.

Paul

Received on Friday, 31 May 1996 19:19:43 UTC