> >... The rule of http headers over-riding embedded self-desciption is going > to be necessary until this has come to pass. > >Even after then it's going to be necessary as there is only one http header > which states encoding, but there is an unlimited number of mechanisms for > self-description in an unlimited number of potential document types. > > > > > This follows only if you accept the principle that the carrier has the > duty to ensure that the recipient can understand what they receive. I > don't accept that the carrier has the duty or even the right to do that. > In fact I would suggest that it is an infringement of my civil liberties > to do so just as much as it would be for the snail mail service to > censor or even reformat my mail.
My webserver has a duty and right to do that if I want it to. Again what is happening in Doug's case is clearly an error on the part of the server, I am only saying that the error is not in the policy of http headers over-riding document self-desciption. Had Doug more control over his server he might want it to do the sort of re-encoding when a browser requested it. In other scenarios the server might be the closest thing there is to an "author", or the author might provide some XML giving the core of the document with the server adding other features such as navigation, records of user comments etc. which puts the server in a far more authoritative position than when used for "straight" file transfer (and much of my bread-and-butter is such systems). In Doug's case a server should act exactly as you say and leave well alone, but that is not the only case that the protocols involved have to serve.
