- From: Adam M. Donahue <adam@cyber-guru.com>
- Date: Wed, 13 May 1998 23:43:00 -0400
- To: http-wg@cuckoo.hpl.hp.com, Dave Kristol <dmk@research.bell-labs.com>
> 4) <http://www.ics.uci.edu/pub/ietf/http/hypermail/1998q2/0069.html> > > I think allowing TE: identity; q=0 is a bad idea. I think "identity" > should always be allowed. I thought Jerrfey Mogul made a good point about wanting to forbid identity when compression would be highly desireable (as in the case where bandwidth is very expensive). However, I agree with you that since chunked encoding is _always_ acceptable it seems somewhat silly to disallow identity in that case. So... Perhaps in some cases the amount of data to send is so large that the receiving computer won't have enough space allocated in its internal representation of the content length to store the Content- Length header's value. ;-) I guess also that if you're thinking of performance, maybe the idea is that the hex representation of chunked transfer is more efficient to convert to binary. And then there's the case of an interrupted transfer. It's probably easier for a client to figure out which characters were received successfully in order to request the rest if those characters come in little chunks rather than one long stream... OK, these are a stretch. Adam Adam mailto:adam@cyber-guru.com
Received on Wednesday, 13 May 1998 20:45:47 UTC