- From: Larry Masinter <masinter@parc.xerox.com>
- Date: Fri, 16 Dec 1994 12:57:06 PST
- To: mogul@pa.dec.com
- Cc: Albert-Lunde@nwu.edu, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
I used to be on the other side of the fence (I thought scanning for boundaries was going to be too expensive), but Ned Freed convinced me. Now I'm a convert. Boundary scanning is also more *reliable* than content-length, because length calculations are unreliable. > I was responding to Mitra's point that if you don't use some sort of > encoding on the data, a file could conceivably contain > <CR><LF>--boundary--<CR><LF> (or whatever else you wanted to use). I can > envision a situation where you couldn't download httpd using http because > the file contained the boundary string. Using content-length to determine > the end of a document would be 100% reliable, since the actual content > couldn't possibly conflict with the protocol information. I only require 99.99999% reliability. (Besides, I'm in the market for a Pentium.)
Received on Friday, 16 December 1994 12:58:30 UTC