Re: Including 'fragment identifier semantics' in MIME media type registration?

> True but insufficient and possibly misleading.
> The reason for having multiple locator types is to
> ensure that one can link into a document where one
> may have read only rights or for example, the format
> is not structured, say a binary image.  HyTime/DSSSL
> designers finally understood that markup is simply
> YetAnotherNotation and made progress on the issue.
> XML is YetAnotherNotation.  The format handler type
> determines the means of resolving the locator but if
> names cannot be supported (what you call an anchor)
> as locators, then other types have to be.
>
> Encouraging names/ids may be good practice, but not much more.
> Linking into a binary image or a strip of film or even
> a position of a book on a library shelf usually requires
> more than a name.  Not new news.  I do understand the
> maintenance issues of positional and chained locators.

I agree that such mechanisms are necessary in some systems; I dispute
whether they should be part of the Web architecture.

Linking where you have read-only rights isn't robust; unlike with the
resource/URI relationship, there is no guidance, much less guarantee, that
a representation's structure is stable over time.

Received on Wednesday, 4 September 2002 18:53:06 UTC