- From: Jeffrey Mogul <Jeff.Mogul@hp.com>
- Date: Thu, 10 Aug 2006 15:34:57 -0700
- To: Julian Reschke <julian.reschke@gmx.de>
- cc: HTTP Working Group <ietf-http-wg@w3.org>
Julian Reschke wrote:
>> Well, that's why I was submitting a draft -- feedback is needed. I
>> realize that Entity-Transform currently may violate the KISS principle,
>> maybe I should go back to it's minimal form of...:
>>
>> Entity-Transform = "Entity-Transform" ":" 1#transform-info
>> transform-info = "identity" | "unspecified"
Hmm, I think this would still need to be
transform-info = "identity" | "unspecified" | extension
extension = token
...so it *can* be extended later.
More feedback appreciated... Do people think it's worthwhile for a
server to be able to specify the kind of transform?
I thought about this a little more. There's a potential problem
with simply specifying the grammar as
Entity-Transform = "Entity-Transform" ":" 1#transform-info
The problem is that this (1) allows multiple transforms, and
(2) doesn't explicitly specify what order they were applied in.
But one could easily imagine that, if a server does apply
multiple transforms, they are not commutative. For example
(not that you would ever expect this combination), "uuencode"
followed by "remove-nasty-words" is not the same as
"remove-nasty-words" followed by "uuencode".
See the language in RFC2616 for Content-Encoding as an example
for how to solve this problem:
If multiple encodings have been applied to an entity, the content
codings MUST be listed in the order in which they were applied.
Without such an explicit requirement, one could imagine
implementations that use other orders (such as: alphabetical,
or reverse-order).
This is another case of where an incautious over-application
of the KISS principle could lead to ambiguity and hence
complexity.
-Jeff
Received on Thursday, 10 August 2006 22:35:12 UTC