[freearchitecture] Presentation to the list and some ideas

[Bruno Postle]

On Sun 21-Sep-2003 at 01:44:04AM +0200, Rafael Villar Burke wrote:
> 
> Your debate about a free CAD format and a CAD program design
> approach catched all my attention, because I think any
> architect/engineer in the free software community feels terribly
> the lack of such tools with professional quality.

This has generated quite a bit of discussion on the cad-linux
mailing-list:

    http://www.freelists.org/archives/cad-linux/09-2003/

Though I'm not sure if it will lead to anything.  Interestingly, it
turns out that the idea of splitting monolithic file formats into
many small CVS-friendly files has been thought-of elsewhere:

    http://xspace.sf.net/

'Xspace' is at almost exactly the same proof-of-concept stage as
'draft', but the implementation couldn't be more different.

> The approach of organising entities as files has already been
> taken by a comercial CAD tool used in Architecture and Urban
> planning called Arris.  This tool appeared several years ago and
> ran on top of a Xenix OS and now M$-NT, but I think nobody else
> has repeated that strategy so far. As you stated it is an easy way
> to allow concurrent editing.

Thanks, I've never used Arris, is it right that it uses one-file-per
layer?

    http://www.arriscad.com/intro2arris.html

> Another tool that came to my mind reading your proposals is LDAP. It is
> a database system implemented with information organised in files and
> directories to represent objects and its structure and relationships.

It could be useful, apparently postgresql has specific datatypes for
geometry which is another interesting possibility.

> Some objections I see in the entity=file approach is that files usually
> organise as B-Trees for searching, and R-Trees or Quad-trees are best
> fit for CAD databases. With large sets of data this could be an
> important issue, but I have no tests nor data about this.

I had a look around, the R-Tree and Quad-Tree stuff seems to be all
about searching data-structures already in memory?  I'd like to
think that a clean file-format wouldn't have such optimisations
built-in.

> Another subject I also think it's important is the XML one. I tend to
> think that the advantages of the format surpass its disadvantages.
> 
> Some reasons are:
> 
> - XML tools are starting to flourish and it will surely be a widely
> supported format. Wouldn't it be great to have these tools for use with
> CAD data?.
> 
> - XML allows validation.
> 
> - As to the lack of space-awareness of the XML format, wouldn't any
> human readable format probably show the same problem?. Furthermore, if
> the tools used to generate the data are shared between users this
> problem should not arise unless someone draws "by hand".

At work the other day, two of my colleagues were both trying to edit
different parts of the same word-processing document (a business
plan).  They asked if it would be possible to put the OpenOffice XML
document into CVS and let that resolve any conflicts, or just send
each other diffs of their changes.

Of course they couldn't, XML doesn't work like that, and people
wonder why I use Vi for writing text...

Actually I was pointed to the PYX spec for an XML-variation with
strict formatting that was designed for use with line-oriented tools
like diff and patch:

    http://www.xml.com/pub/a/2000/03/15/feature/index.html

..so maybe that is a possibility too.

-- 
Bruno