> Anders W. Tell writes:
>
> > So adding an attributes at the end allows the developers to keep a
> > pure stream programming model without sacrificing performance.
>
> Yes, but what you save in the writing, you lose in the reading: a SAX
> client reading your XML element would have to buffer the whole
> contents until it got to the end tag to make certain that there were
> no important attributes there.
Yes and No, Since end-tag attributes is an extension to current xml the above
problem may
be offset by adding an end-tag marker to attributes in the DTD, DCD, XSchema's.
Which means the applications know that the specific attribute may/must occur in
the end-tag.
>
> A specific version of a specific XML document is written only once,
> but it can be read many times; when you are forced to choose between
> adding complexity to writing or reading, writing should end up drawing
> the shorter straw.
Depends on the UseCase in question, certainly on the internet there is probably a
80-20 (read-write)
balance for *static* documents but more and more application/servers generates
documents or fragments
on demand with no possibility to cache. An example of this are dynamic queries.
Another UseCase is InterProcessCommunication such as WebRPC. Here *documents* are
generated
on the fly and discarded on reception and caching many documents/requests/replies
may not be a viable solution.
As far as complexity goes, the only change is to add StartTag APIs to EndTag's.
Im not proposing to add end-tag attributes, just trying to understand the
implications if the feature was available..
Best,
/anders
-- /_/_/_/_/_/_/_/_/_/_/_/_/_/_/ / Financial Toolsmiths AB / / Anders W. Tell / /_/_/_/_/_/_/_/_/_/_/_/_/_/_/