Re: Byte ranges -- formal spec proposal

>Why should your server ever receive them, and why shouldn't it just
>ignore them? 

Well, I don't control clients. Dynaweb is somewhat unusual in that it
does not serve from a filesystem directly, but rather, from a
database, and it relies heavily upon data structure. If people are led
to believe that byte ranges are accepted (because most filesystem
based servers will support them), then they will necessarily expect
something which claims "to, for all appearances, be just another HTTP
server", to also support them (as a cheap way to take a peek at a
document of whatever).

Given that users expect them to be supported, they'll be surprised
when a request for 500 bytes returns 50000 instead, or if the server
returns an error code. They'll be more surprised to find that each
time they request the same 500 bytes, they could, possibly, get 50000
*different* bytes back. 

There are inumerable failure cases for this scheme; inumerable cases
where it is meaningless. It seems rather silly to pollute the entire
URL namespace with application-specific extensions. If we limit
ourselves to just HTTP URL's, then a new method makes better sense,
because we can at least respond in a reasonable manner.

I notice that Larry has said the same thing as well.

Received on Thursday, 18 May 1995 08:01:30 UTC