Re: gEDA-dev: [pcb] physical layer stack

[Peter Clifton <pcjc2@xxxxxxxxx>]

On Tue, 2008-02-19 at 00:12 +0100, Jakub Piotr Cłapa wrote:
> Peter Clifton wrote:
> > The real lesson I think we could learn from XML is decent name-spacing
> > for extensibility.
> 
> But it's not free and done. You still need to parse this, save in memory 
> and write back to disk when saving. And this means you will likely need 
> to use the XML tree as your internal data structures. I'm not sure if 
> namespaces not consciously supported by PCB source code will be used for 
> anything.

PCB supports plugins, so there is a reason to want namespaced data in
the file.

> There is also the thing about backward compatibility and ease 
> of extension but once more the semantics of supporting older versions 
> with newer files are something you must do; not something you get for 
> free from XML.

The view that XML will cure the world and avoid the need for
data-structure design is an idea which has harmed XML lots, and one I'm
trying to avoid bringing across here.

>  In this regard XML with namespaces is not something way 
> better than any other resonable format (like s-exprs with a 
> deterministic writer or netstrings).

If you namespace the s-exprs, its another valid syntax. Deterministic
writing is the key for any more complex format.

> The only place where I have seen namespaces really shine was Jabber/XMPP 
> because the produced data got consumed immediately and the win was that 
> you could ignore what you didn't understand (like with HTML) while 
> validating the rest (unlike HTML).

The amount of bad HTML browsers accept was IMO a stupid design mistake
in HTML. Far better is defining a standard syntax (allowing expansion),
standard semantics, and dropping a big fat "ERROR" page if the syntax or
defined semantics are violated. There would be fewer bad web-pages out
there if this were so!

[snip]

> > You can still happily write out an XML file with printfs, awk, etc..,
> > but to then re-read a file, you would need to either:
> > 
> > a) Use a proper XML parser
> > b) Have some guarantees from PCB about the white-space formatting it
> > uses when saving.
> > 
> > a) isn't totally abhorrent, if you assume this parser lives in PCB, and
> > PCB were to expose a swiss-army-knife's worth of actions for scripting
> > with.
> > 
> > As DJ said, at some point we may need to go multi-line, and that means
> > more parsing complexity. In any case, we need to "invent" this new
> > format from a data-structures POV. Regarding syntax, if not XML, then
> > we'd need to invent our own syntax and parsers too.
> 
> A is reasonable and B is probably a good idea. I think that A will be 
> good for serious tools while B will allow one-liners. Of course the XML 
> as base means we may trivialy map it to s-expr or any other reasonable 
> data format (is't just trees; nested lists).
> 
> Why do you think you will need to go multiline? Probably one line per 
> element/line/pad would not be a problem. And for easier inspection you 
> could indent it with xmlint or even some GUI XML editor.

I haven't done a data-structure design for PCB or any other layout code,
however I got the feeling the amount of information which "might" be
added could potentially be quite large.

-- 
Peter Clifton

Electrical Engineering Division,
Engineering Department,
University of Cambridge,
9, JJ Thomson Avenue,
Cambridge
CB3 0FA

Tel: +44 (0)7729 980173 - (No signal in the lab!)



_______________________________________________
geda-dev mailing list
geda-dev@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-dev