See below...
Victor Mote wrote:
Peter B. West wrote:
As to the necessary conditions for committer status, "Shall we take that as read, darling?"* The question remains, "If a another developer happens along who 1) is persuaded that alt.design is worthwhile, and 2) sees the existing properties code as a working implementation that is better, and 3) wants to work on alt.design more or less exclusively, will he/she be admitted to committer status with the - all other things being equal - now customary alacrity?"
Will those existing committers who are not interested in alt.design allow it to flourish in the (unlikely) event that it attracts the interest of other developers, or will the Party line, necessary as that may be considered, dictate the such a development be resisted?
I have exactly one vote on such matters, so I can't speak for the whole, but as far as I am concerned, developers such as you describe are more than welcome to join the party. In the event that alt-design remains on a branch, I don't think any reasonable person could object. At the point in time that we contemplate merging to the trunk, we need to come to an agreement. In the unlikely event that we can't come to an agreement, we always have the option to fork the project. My purpose here is to avoid that if possible.
BTW, I hope this isn't a Peter-vs.-Victor thing.
Absolutely not. This is a question for all of the existing committers. Your response above is exactly the one I would hope to see from everyone. Incidentally, the project is already de facto forked - the purpose of integration is to bring the benefits of alt.design into the main redesign.
For example, I know there are opportunities to use less memory and more speed (which you report in alt-design) in the FO tree creation. If memory serves, we are storing the URL for the fo: namespace in every FONode object, which should be replaced by an integer pointing into a table. I am very open to being educated, but I think it is fair to say that I am not persuaded on all of it yet, and I think the burden of proof lies heavily on you. Unless pull parsing is an integral part of the whole, I think the alt-design changes will be best swallowed in smaller pieces.
alt.design is a complete rewrite of the front end. It could possibly be broken up and rejigged to use a standard SAX approach, but that would require major surgery, and would obviate a great deal of the advantage of doing away with the design convolution which is imposed by SAX.
There is no great design rationale in the basics of the current approach to FO tree parsing. It is that way because it is obliged to be by SAX. SAX imposes itself on design, unless it is forcibly restrained, and that imposition will, IMO, always be to the detriment of the design of complex systems with a well-understood structure to their data.
Clearly, that is a view that is not shared here, but it is one, which, I believe, has been and is being debated at length elsewhere. Andy Clark's NekoPull variations on Xerces XNI are an attempt to provide a pull alternative for Open Source parsing. It seems unproductive to try to debate the issue here; one either takes to it or one doesn't.
Also, if pull parsing can be implemented in a pluggable, configurable way (ie. choose either push parsing or pull parsing at runtime), it will have a much better chance of being accepted. My understanding of it is that it has a pervasive effect, making separation between the modules more difficult instead of less. IMO, absent a /significant/ benefit that cannot be achieved some other way, this would be a deal-killer.
Which modules? I'm not sure what you mean. The modularity of area processing? The Rec gives an utterly spurious view of this. It has been misleading developers since the drafts were first published, and giving the impression that things can be neatly modularised. Some of the persistent difficulties of design arise from this misunderstanding. However, I may be misreading you here. What modules do you have in mind?
Peter -- Peter B. West http://www.powerup.com.au/~pbwest/resume.html
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, email: [EMAIL PROTECTED]
