Re: ERB call on addressing

On Mar 27,  2:26pm, lee@sq.com wrote:
> Subject: Re: ERB call on addressing
> Matthew Fuchs <matt@wdi.disney.com> wrote:
> > It seems to me the essential question is, if I have separate URLs
> > for each chapter, then they can refer to each other, but if I have
> > one URL for the book, and retrieve each chapter as a query, then any
> > cross references require returning to the server, because each cross
> > reference is a query.
> If you download the whole book, you don't need to go back to the
> server tofind fragments.


> If you only download a fragment, the fragmenter can change links
> in the document so that they point to the correct fragment.

Yes, you can alpha rename everything.  This is _not_ a no brainer.

> There is nothing but a URL.  Every URL is a query -- it's just that
> sometimes the remote server answers the query by sending a disk file.

Exactly what I said in the paragraph you cut.

> A URL like http://www.softquad.com/sgml/ might be sending back a file
> called index.html or default.htm or it might be looking in a database,
> but you (the client) can't tell.
> > So we want a syntax that says "give me piece X
> > of document Y".
> You already have it.
> 	http://server/documentY/partX
> might do this, if the server is so set up.  Jon Bosak has illustrated
> the use of this syntax for us already, has a working server that does it,
> and did not need anything special at the server end.

Yes, this can obviously be done.  But the client has no way to know that
http://server/documentY/, http://server/documentY/partX and
http://server/documentY/partY have any relationship because it's only "might."
In most of the web they don't.  I don't think renaming is so easy once you
start thinking through examples, i.e., suppose subZ of partX can be separately
downloaded, and I get partY, which has a reference to
http://server/documentY/partX/subZ (because after all, I could ask for just
that), but then I request partX and a little later subZ.  With a little thought
it's pretty likely they are the same, but there is no guarantee.  I now have
two versions of subZ.  This is not good behavior.  I think it's worthwhile for
the client to know both subZs are more than coincidentally the same.

Matthew Fuchs


Follow-Ups: References: