- From: M. Hedlund <hedlund@best.com>
- Date: Sat, 27 Jan 1996 10:43:10 -0700
- To: BearHeart / Bill Weinman <bearheart@bearnet.com>, Brian Behlendorf <brian@organic.com>, www-talk@w3.org
At 12:37 AM 1/27/96, BearHeart / Bill Weinman wrote: > >At 10:11 pm 1/26/96 -0800, Brian Behlendorf spake: >>Maybe we should thank Microsoft for pointing out how silly and utterly >>useless User-Agent is for content negotiation. [...] > > User-Agent may not be the most technologically whiz-bang thing >you can think of for content-negotiation, but it works. > > Your justifying Microsoft's mockery of it is a slap-in-the-face to >all the people who are working their collective ass off in a monumental, >cooperative, and VOLUNTEER effort to create a set of standards that >will allow the Net to grow and thrive. >[...] > I thank them for nothing, > and I thank you for nothing. Brian is at a WWW standards meeting in Boston right now (oh, the irony), and therefore can't respond. Allow me. User-agent negotiation _does not_ work. Some feature X is implemented in User-Agent Y. The erroneous assumption made by user-agent negotiation is that all browsers not-Y do not have feature X. Microsoft, implementing feature X in browser not-Y, was unable to make use of that feature, and therefore posed as browser Y. If user-agent negotiation did work, as soon as MSIE hit the streets, all negotiating sites would immediately add MSIE to the list of X-capable browsers used in their negotiation algorithm. Of course, that doesn't happen. A real negotiation solution would allow a browser to advertise its capabilities automatically and concisely, saying "I can do feature X."[1] Microsoft's solution: 1) works for them; and 2) demonstrates the frustration non-Netscape developers feel with this situation. (Maybe Netscape developers, too, but it seems less of a problem for them.) What I found silly about MSIE 2.0b for Mac was that, by default, it "claims" to implement features that it doesn't [2]. What I find irritating is that a standards body exists to try to solve this problem [3], and Microsoft hasn't decided that this is enough of a problem to send a representative. (If anyone associated with MSIE is interested in joining, send me mail and I'll reply with group info.) These are all just data points intended to drive the development of a real negotiation solution. Marc Hedlund <hedlund@best.com>, <marc@organic.com> [1] Another solution would be to devise a conditional HTML system that allowed the browser to make rendering choices based on feature-identifiers. Netscape's <frameset>/<noframes> tags are an instance of this, as is the "fallback text" in the <applet> tag. For that matter, the "alt" parameter to the <img> tag is also a form of conditional HTML. The problem is that there is a different solution for each of the three tags I just mentioned -- there's no generalized system. Brian and I have been working on a document proposing such a system. [2] I received mail after my last message saying the MSIE 2.0B for Mac does indeed render server-pushes. If this is the case, my apologies for the error. I tested MSIE on <URL:http://www.levi.com/menu.nhtml> and got a bunch of broken-image icons. [3] Bill, your slam on Brian for his "slap-in-the-face" to those "working their collective ass off in a monumental, cooperative, and VOLUNTEER effort to create a set of standards" is ill-informed (not to mention overwrought). Brian is one of the ten or so members of the content negotiation subgroup. He is one of the people you claim he is insulting.
Received on Saturday, 27 January 1996 13:42:02 UTC