Re: [css21][css3][svg] SVG and unit-less length values

Coming back to the topic and to resume:

1. There is an obvious requirement for SVG document
authors to use no units within the content of an SVG document,
a) they do not need units identifiers for local units, practically it is
    a unit of '1', no need to define another explicit symbol for '1' or to
    write something like '*1 ' instead of ' '.
b) units identifiers blow up the source code unnecessarily
c) the relation between absolute units, px and em is not predictable
    an not implemented completely correct in any viewer I tested
d) The use case CSS+(X)HTML and the current unforunate problems
    of the CSS-WG with units show, that they are not really usable as
    intented and as what they are - if a cm cannot be displayed as a cm
    in a viewer, either the unit or the viewer is practically useless.
e) why to require something, that does not work properly currently (and
    for the past 10 years) in CSS+(X)HTML and caused obviously an
    infinite discussion and bizarre conclusions?

2. There are a few use cases (already mentioned, technical drawings,
maps etc), that require absolute units, at least absolute units for width
and height of the root svg element. Primarily the lengths have to fit
precisely for printers for example to give such a technical drawing to a
mechanical work shop to produce a real object, else SVG is not usable 
for such applications and authors need still another format or still have to 
draw everthing manually on millimetre paper without the option to reuse and
to archive this on their computers - and will maybe never switch to a 
standard like SVG, if such W3C formats continue to have problems with 
such simple 'real use cases'. 

This is what really happens every day in my job - I have to ask:
Is SVG advanced enough for technical drawings? For scientific graphs?
Typical answer currently is: No! Continue to use another format and
provide only fun documents with SVG.
Is CSS+XHTML (+SVG+MathML) advanced enough to write a scientific
paper? Are the viewers advanced enough to display it in a reliable way?
Typical answer currently is: No! Continue to use LaTex and postscript.
Use CSS+XHTML only for fun documents on the web.

3. To get absolute units approximately right on monitors is technically
possible in many circumstances. Obviously for these situations authors
and the audience should not be cheated by stupid programs and
stupid recommendations.
If such an information is not available, a program has to guess.
Of course a technical recommendation can contain a suggestion
about a resonable guess for different display medias.
Advanced programs may provide an interface on demand to allow 
users to improve the guess due to their own information about the
issue (to improve the 'user experience', if required).

4. One has to take into account, that some authors have not the
capabilities to use a larger amount of different units for different
purposes. There are obviously a lot of stylesheets for CSS+(X)HTML
out there, which document these author problems.
This is a clear indication, that SVG should continue and enforce
efforts to suggest authors to use as less unit identifiers as possible 
in their documents to avoid unnecessary complications.

Because there will be always some authors, that will create bugs
without beeing able to fixing these themselves, the only option
to improve this situation I can see is to provide more information
about 'best practice' and 'how to use and mix units for different
applications properly'.

5. It is not the primary task of a technical recommendation to
fix bugs of authors, especially not if this prevents other authors
from creating documents in a straight forward way for their own use 
cases without bugs.
If a technical recommendation focuses too much on the incompetent,
well, then it becomes a recommendation only for the incompentent
and advanced authors have to use a better format to realise their
use cases. They have these use cases, this cannot be changed with
a technical recommendation or an infinite discussion about the
question, if people of this mailing list have such a use case or not.


Received on Sunday, 15 August 2010 11:05:26 UTC