- From: Ian Duncan <id@cc.mcgill.ca>
- Date: Thu, 8 Feb 1996 11:44:22 -0500 (EST)
- To: hallam@w3.org
- Cc: hallam@w3.org, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
On Wed, 7 Feb 1996 hallam@w3.org wrote: > Imagine that we use HTTP to create a "hyperterminal" type application. Ie we > connect to the resource and it spits out a continuous stream of data in an IRC > type manner. I cannot imagine ... > Now imagine we use this feature to create a "debugging port" on a server. With several well-designed protocols ambient to handle this particular application anyone who jams HTTP around the problem, I really might enjoy imagining fates worse than simple predictable failure. Put another way, there are a thousand ports in the net and each should offer visitors the best from their particular domain. Those who wish to turn HTTP into container for all that litters the landscape wish for a sack of trash. > Now imagine we use the port to debug the connection made to the port. The MIME > boundary is bounced back as data rather than as control. Result unhapiness. I can imagine some useful circumstance that might loopback a MIME boundary marker in a streaming HTTP connection but making the second loopback without understanding and provisioning a filter, with a capability to escape the marker, would be commiting a serious design error, likely with other unintended losing consequence. > My college tutor at Oxford would be unhappy were I to go round introducing > unnecessary bug potential even if the probability of the bug occuring were > marginal. Mucks up a proof of correctness no end... And I'd thought with all those fine modern physicists working there that Oxford would have finally left 19th C. methods of modelling the universe behind. As Ned Freed and Jeff Mogul generously explained, unintentionally spitting out any very large random number is significantly smaller than other more harmful sources of noise in the system. /id
Received on Thursday, 8 February 1996 08:47:48 UTC