[cups.general] Error: No %%BoundingBox: comment in header!

[Michael Sweet]

This discussion is getting nowhere, since you aren't listening.  I will
respond one more time, with the preface that we have no plans of
changing how CUPS works WRT needing the BoundingBox to actually support
things like number-up and fitplot.

Ambrose Li wrote:
> ...
> Which means:
> 
> - Apple is not wrong (it did not violate the correct specification which is
>  DSC and not EPSF)

Apple technically is not wrong, but they will in fact be adding the
bounding box comments in a future version of OS X because they want
things to work...

> - CUPS' reliance on %%BoundingBox rests on an assumption (that the
>  bounding box must correlate to the page size) which is not necessarily
>  true

Perhaps, however we have no other way to find out the document size.
In the absence of this information, we have to depend on the media
size specified for the job.

Also, we don't depend on BoundingBox to determine the page size,
we only use it when *scaling* the page - basically for the number-up
and fitplot options, where you *expect* that the page will be scaled.
In the real world, this means that the page margins are "discarded"
when you print N-up.

> - DSC-compliant jobs without %%BoundingBox exist in the real world
>  and so on a practical side should be considered

They are, but there are limitations - you have to have the same
input and output size, for example.  BTW, the *only* options
that are affected by this are fitplot and number-up - everything
else works on the output page size dimensions.

> This is something our views about what CUPS should be diverges:
> IMHO CUPS should always be able to emulate "no spooler".

Well, feel free to develop your own print system that does this.
The rest of the world is moving away from PostScript as a spool
format because of all these problems...

> ...
> The fact that Apple now generate PostScript that actually conforms to
> the DSC and that CUPS is not able to (or claims to be not able to) handle
> properly points to an interoperability problem with CUPS. These are not
> my "complex requirements", but a real-world situation that is valid and
> will be occuring increasingly often.

First, there have long been issues (which Apple is still resolving)
with the DSC conformance of Apple's PS output.  BoundingBox is only
one (very small) issue, there are still lines longer than 255
characters, embedded binary data without the corresponding comments,
and incorrect comment data which can lead to problems when printing.

We go out of our way to support basic printing with any PS file, but
if you expect the file to be unmodified by CUPS, you have to configure
it that way and accept that you will not have access to any of the
features that CUPS provides.  You'll essentially have the old LPD or
LP print spoolers which have been left behind *because* they cannot
support more advanced printing.

Similarly, if you want access to CUPS features, you need to provide
CUPS with the information it needs to support them.  If you don't
provide the necessary information, you have to accept that the guesses
it makes on your behalf will not always be right.

Whether BoundingBox is the "right" information for number-up and
fitplot is not up for debate - it is the *only* standard comment that
can provide reliable information on the printed area of the page.  No,
DocumentMedia is not widely used or consistently implemented, so the
information you get there is not reliable!

-- 
______________________________________________________________________
Michael Sweet, Easy Software Products           mike at easysw dot com
Internet Printing and Publishing Software        http://www.easysw.com