RE: revised CGZ proposal

At 09:50 PM 12/4/2008 -0500, Weidenbrueck, Dieter wrote:
>I agree with this use case.
>
>  >  2.) Add to the end of @@7.1:  "WebCGM 2.1 viewers, both static and
>  >  dynamic, must correctly handle valid WebCGM 2.1 Binary-encoded metafiles
>  >  that are gzip-compressed."
>This statement talks about WebCGM 2.1 viewers (it should rather be 
>interpreters IMO, how else would IsoDraw be able to read cgz files 
>otherwise?),

Maybe so.  I used "viewers" because that is the way Ch.7 is currently 
written.  We would need to introduce "interpreter" and define it, I would 
think, if it becomes part of a conformance definition.

>and even attaches a version number to the viewer itself. So to clarify: we 
>are defining the profile version WebCGM 2.1. Is a WebCGM 2.1 interpreter 
>one that is can read compliant WebCGM 2.1 files, or is it an interpreter 
>version 2.1 that can read WebCGM 2.1 files in zipped and unzipped form?

The latter.  An interpreter that is webcgm 2.1 conformant can correctly 
handle valid WebCGM 2.1 instances in both zipped and unzipped form.

>
>If the latter is the case, then we can say, that we introduced a version 
>number for interpreters and their capabilities,

Perhaps I'm misunderstanding your point, but... It has always been so in 
WebCGM.  See Ch.7.  "Viewer" is a Class of Product for conformance 
purposes, and each version of the WebCGM defines (throughout the 
specification) conformance criteria for each Class of Product, for that 
version.  I.e., each version x.y of WebCGM so far defines an x.y-compliant 
viewer.

Example.  A 2.1 compliant viewer is not necessarily 1.0 compliant.  It 
would NOT be, for example, if it didn't handle multi-picture metafiles 
(which were a feature of 1.0, then were deprecated and removed.)

>which means, that compliant interpreters must be able to read zip 
>compressed CGM files from the interpreter version 2.1 on. This doesn't say 
>anything about the profile or the profile version of the CGM that they 
>would be able to read.

I don't understand this point -- "from the interpreter version 2.1 on"?  (I 
have some suspicion that it may conflict with conformance definitions and 
the deprecation and obsoletion concepts of ch.7.)

>
>Probably this is too subtile a difference, but if we treat interpreters 
>this way it would work for any kind of CGM file to be compressed.

...too subtle, yes -- I think I'm missing your point.

-Lofton.


>
>----------
>From: public-webcgm-wg-request@w3.org 
>[mailto:public-webcgm-wg-request@w3.org] On Behalf Of David Cruikshank
>Sent: Donnerstag, 4. Dezember 2008 20:17
>To: Lofton Henderson
>Cc: WebCGM WG
>Subject: Re: revised CGZ proposal
>
>Good, the classic use case, as I described this morning, is that I've got 
>a database of thousands of cgm files.  I'm going to implement some WebCGM 
>2.1 functionality in a few of them and want to take advantage of gzip 
>compression for all the files in the database.  I do not want to touch 
>each cgm file to update its ProfileEd to 2.1, just to make us of gzip.
>
>Dave
>
>On Thu, Dec 4, 2008 at 10:56 AM, Lofton Henderson 
><<mailto:lofton@rockynet.com>lofton@rockynet.com> wrote:
>Thanks for the comments, Don.
>
>Per today's telecon, Part 1 is closed and agreed.
>
>Part 2, the "lurking question", is what we're seeking comment on.  You 
>opined, "...a WebCGM 2.1 viewer should handle legacy WebCGM gziped 
>files".  To be precise, the requirement would be phrased for 2.1 viewers 
>that claim 1.0 and/or 2.0 backward compatibility.  I.e., if they claim 
>conformant handling of 1.0 and/or 2.0 metafiles, then they shall correctly 
>handle gzip'd versions of such.
>
>I would imagine this includes all 2.1 viewers in practice, but the wording 
>needs to be careful, because a viewer can in theory be 2.1 conformant but 
>not 1.0 conformant.
>
>(FWIW, I agree that it is a sensible and practical requirement.)
>-Lofton.
>
>
>
>
>
>At 11:21 AM 12/4/2008 -0600, Don wrote:
>
>Lofton,
>
>  >  All --
>
>  >  I'm going to propose a slightly different resolution than what is in the
>  >  previous (below-attached) thread.  Approval of this change will be on 
> the
>  >  next WG telecon agenda (04-dec).
>
>  >  Part 1:  basic resolution
>  >  ==========
>
>  >  Basically, to accommodate dissatisfaction with potential bending the
>  >  CGM:1999 Rules for Profiles, we will instead make gzip support a 
> Chapter 7
>  >  viewer requirement, that WebCGM 2.1 viewers properly handle
>  >  gzip-compressed Binary-encoded WebCGM 2.1 content.
>  >  The proposal:
>
>  >  1.) remove the "Other:..." gzip requirement from T.13.1 in Chapter 6;
>  >  2.) Add to the end of @@7.1:  "WebCGM 2.1 viewers, both static and
>  >  dynamic, must correctly handle valid WebCGM 2.1 Binary-encoded metafiles
>  >  that are gzip-compressed."
>
>  >  That, I believe, captures the sense of recent discussions.  Comments on
>  >  that part of it?
>
>I agree.
>
>  >  Part 2:  a lurking question
>  >  ===========
>
>  >  I will point out in advance that this revision leaves open a question
>  >  about WebCGM 2.1 viewers and gzip'd 2.0 or 1.0 content.  If you look at
>  >  Ch.7's deprecation and obsoletion wording, for example, you will see 
> that
>  >  we don't require that a conforming 2.1 viewer be a conforming 2.0 viewer
>  >  (nor 1.0).  But 2.1 viewers can optionally feature 2.0/1.0 
> conformance --
>  >  i.e., correct handling of 2.0 and 1.0 metafiles per the 2.0/1.0
>  >  conformance rules.  In fact, I would guess that almost all are built 
> that
>  >  way.
>
>  >  So the question:  do we want to apply the 2.1 viewer gzip 
> requirements to
>  >  older content (2.0/1.0), for those 2.1 viewers that claim backward
>  >  compatibility for 2.0/1.0?  I.e., something like "In addition, 2.1 
> viewers
>  >  that correctly handle valid 2.0 or 1.0 content per the conformance
>  >  requirements of those (2.0 or 1.0) WebCGM versions shall correctly 
> handle
>  >  such metafiles when they are gzip compressed."
>
>  >  Please comment on that refinement, pro or con.  What do people want?
>
>I think that a WebCGM 2.1 viewer should handle legacy WebCGM gziped files.
>Since the viewer needs to un-gzip to determine the profile anyway, the viewer
>might as well go ahead and handle the un-gzipped contents.
>
>Don.
>
>  >  (In parallel, Lofton has an action item to contact Dick Puk at SC24 
> about
>  >  processing a corrigendum or addendum to ISO 8632, that would add
>  >  gzip-compressed Binary, at least, to the valid encoding/content types of
>  >  ISO 8632).
>  >  -Lofton.
>
>  >  At 02:54 PM 11/12/2008 -0800, David Cruikshank wrote:
>  >  I like this solution.  As long as I'm using a 2.1 viewer, it doesn't
>  >  prevent me from using zipped cgm files in my database of graphics.
>
>  >  Dave
>
>  >  On Wed, Nov 12, 2008 at 12:13 PM, Lofton Henderson 
> <<mailto:lofton@rockynet.com>lofton@rockynet.com>
>  >  wrote:
>  >  All --
>
>  >  Having read preferences and concerns in this thread, and following on 
> from
>  >  the discussion that happened this morning in the TC's telecon, IMHO this
>  >  is the right way to handle the CGZ questions that Benoit raised...
>
>  >  1.) completely clarify in 2.1 text that gzip'd binary CGM (content) is a
>  >  valid conformance "Class of Product" for WebCGM 2.1, and that all 2.1
>  >  interpreters must handle it.
>
>  >  2.) say *nothing* in the 2.1 spec about 1.0, 2.0, or other 
> profiles.  (In
>  >  other words, use of gzip'd binary CGM in other contexts is by private
>  >  agreement about archival and transport, and there will be no suggestion,
>  >  MUST, SHOULD, or whatever -- neither implied nor explicit -- in the 2.1
>  >  spec about it.)
>
>  >  This seemed agreeable to everyone this morning, and seems to resolve
>  >  concerns about creating retroactive, even implied, requirements about
>  >  already deployed systems for 1.0, 2.0, and other profiles.
>  >  Are there any objections to this resolution?
>
>  >  Other comments?
>
>  >  Regards,
>  >  -Lofton.
>
>  >
>
>  >  At 03:30 AM 11/12/2008 -0500, Weidenbrueck, Dieter wrote:
>  >  Dave,
>  >
>  >  this is pretty much how I feel about this as well.
>  >
>  >  The challenge is, if we do #1 inside the WebCGM 2.1 profile, then
>  >
>  >  - an unzipped PIP or GREXCHANGE 2.6 file is compliant to the respective
>  >  profile
>  >  - a zipped PIP or GREXCHANGE 2.6 file is compliant to ...what??
>  >
>  >  If a compliant interpreter can read a GREX 2.6 file in zipped form, is
>  >  this a compliant behavior? If it can't, is the interpreter not compliant
>  >  any more?
>  >
>  >  if we set up a rule within a specific profile, it will be valid only for
>  >  that profile. In fact, there is an open question that I can't answer 
> here
>  >  on the road, which is:
>  >  Every CGM file compliant to a profile must be a legal ISO8632 CGM. Is a
>  >  zipped CGM a legal ISO8632 CGM? If you read it following the ISO 
> standard,
>  >  you will have to reject the file as non-CGM after the first bytes. 
> Or, in
>  >  other words, isit legal to define zip compression inside a profile, or
>  >  does it have to be a separate encoding?
>  >
>  >  Dieter
>  >  From: 
> <mailto:public-webcgm-wg-request@w3.org>public-webcgm-wg-request@w3.org
>  >  [mailto:public-webcgm-wg-request@w3.org] On Behalf Of David Cruikshank
>  >  Sent: Dienstag, 11. November 2008 18:31
>  >  To: Lofton Henderson
>  >  Cc: WebCGM WG
>  >  Subject: Re: CGZ files
>  >
>  >  >From a user standpoint, I would lean towards option 1.  That is not
>  >  requiring that cgz only be used with 2.1, but using informative language
>  >  to indicate that cgz files might contain cgm files from previous 
> profiles.
>  >  My reasoning is based on a question I get all the time from people in 
> the
>  >  industry.  "Do I have to open up every CGM file and change the ProfileEd
>  >  every time the industry profile rolls."  Creating a zip content for a 
> CGM
>  >  file can be done without ever touching the file with a CGM tool and I 
> can
>  >  see where an application might take advantage of the cgz file just by
>  >  converting it's whole database.
>  >  A second question....Does anyone really thing a viewing application is
>  >  going to access a cgz file, unzip it, and the if the ProfileEd is not 
> 2.1,
>  >  reject the fiile and not display it?
>
>  >  thx...Dave
>
>  >  On Mon, Nov 10, 2008 at 12:34 PM, Lofton Henderson 
> <<mailto:lofton@rockynet.com>lofton@rockynet.com>
>  >  wrote:  Hi All,
>
>  >  As I'm working on the summary of LC comments, I realize that we haven't
>  >  completely put the CGZ question to rest yet -- there are loose 
> ends.  (See
>  >  my summary below attached.)  Benoit mentioned this also at the end of 
> the
>  >  WG telecon.
>
>  >  About 2.1:  I think we agree that it is a valid content type for 2.1,
>  >  right?  I.e., gzip-compressed binary CGM content is a conforming 
> Class of
>  >  Product for 2.1.  And 2.1 interpreters must handle it.  So we can 
> clarify
>  >  the wording about 2.1 in the 2.1 spec if anyone thinks it is unclear.
>
>  >  About 1.0 & 2.0:  I sense that people want to be able to exchange
>  >  gzip-compressed 1.0 and 2.0 as well.  Correct?
>
>  >  So what are the options:
>
>  >  1.) We could put a non-normative note in 2.1, acknowledging that
>  >  gzip-compressed binary is not conforming 1.0/2.0 content strictly
>  >  speaking, but that it is a useful and used technique, and recommending
>  >  that interpreters should be prepared to handle it.  (And they might
>  >  encounter it in the context of *any* profile of CGM.)
>
>  >  2.) We could process errata for 1.0 & 2.0, making it a formal 
> requirement.
>  >  (This would be a *substantive* errata.)
>
>  >  3.) Other.  E.g. something along the lines of the "storage/transport
>  >  variant outside of CGM".  My question about this option would be:  how
>  >  exactly would such a perspective be reflected in the 2.1 text?  (Is this
>  >  option essentially the same as #1?)
>
>  >  Regards,  -Lofton.
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >  At 08:59 AM 10/21/2008 -0600, Lofton Henderson wrote:
>
>  >  At 12:50 PM 10/20/2008 -0400, Weidenbrueck, Dieter wrote:  [...]  I 
> agree
>  >  with your analysis.
>
>  >  One additional question:  Would it make sense to establish cgz as an
>  >  additional encoding (or  something similar) to make it available to 
> other
>  >  profiles?
>
>  >
>  >  It is an interesting idea, and I hear that people would like to be 
> able to
>  >  do it.
>
>  >  "..establish as an additional encoding.." -- That approach would most
>  >  sensibly involve working a corrigendum to CGM:1999 through ISO.  This
>  >  would not be hard, process-wise.  But on the other hand, CGZ would be a
>  >  somewhat different type of encoding than Binary Encoding or Clear Text
>  >  Encoding.  (It is binary-encoded, then compressed as a whole -- not 
> really
>  >  the same sort of thing as the other encodings, IMHO.)  I'm not  thinking
>  >  about WebCGM 1.0 per se, but other profiles building on ISO 8632 might
>  >  want to use zipping as well.
>
>  >
>  >  Yes, there does seem to be a use case, as I gather from you and from 
> Don.
>  >  I guess in general I find it a shame to restrict usage of zipping to
>  >  WebCGM 2.1 only, and I would like to find a way to be able to use
>  >  compression with any variant of CGM. Probably this could be declared 
> as  a
>  >  kind of a storage/transportation variant outside of CGM?
>  >
>  >  We should be clear about the current situation.  It does not prohibit 
> the
>  >  use of CGZ compression for WebCGM 1.0 instances.  But it isn't supported
>  >  it as a conformance scenario for 1.0 either, i.e., gzip-compressed 
> CGM is
>  >  not a conformance "Class of Product", and handling gzip-compressed 
> CGM is
>  >  not a requirement for the viewer "Class of Product".  (And 1.0 would in
>  >  fact prohibit it, if it were claimed that the compressed instance is 
> valid
>  >  WebCGM 1.0).
>
>  >  What I mean is this... If "BE21" (BE10) is a valid WebCGM 2.1 (1.0) 
> binary
>  >  metafile, CGZ21 (CGZ10) is a gzip-encoded valid BE21 (BE10) metafile, 
> V21
>  >  (V10) is a conforming 2.1 (1.0) viewer, and Z is a standalone
>  >  gzip-decompressor, then...
>
>  >  Valid WebCGM 2.1 scenario:  ...--> CGZ21 --> [V21]
>
>  >  Valid WebCGM 1.0 scenario:  ...-->CGZ10 ---> [Z] --> BE10 --> [V10]
>
>  >  That is, nothing prevents gzip compression of WebCGM 1.0 metafiles.  But
>  >  it (CGZ10) has no standing as a valid content type for a V10 viewer
>  >  according to the WebCGM 2.1 spec's conformance section.  (Note that the
>  >  V21 viewer implicitly or logically integrates the process "Z", 
> because of
>  >  the conformance specifications of WebCGM 2.1).
>
>  >  This is kind of like what Dieter says -- that it is handled as part 
> of the
>  >  transport process, outside of the specifications of the WebCGM 1.0
>  >  specification.
>
>  >  So ... where do we go from here?  Should we put something (informative)
>  >  into 2.1, pointing out that the conforming 2.1 scenario can 
> essentially be
>  >  achieved also with 1.0 metafiles, except the compressed content CGZ10 is
>  >  not a conforming1.0 "class of product", and therefore must be 
> decompressed
>  >  into a valid BE10 metafile before it is a conforming input to a V10 
> viewer
>  >  process?
>
>  >  Should we process errata for WebCGM 1.0 and WebCGM 2.0?  (Difficult,
>  >  because can be argued as a substantive/technical change, and OASIS 
> process
>  >  prohibits such in Errata.)
>
>  >  Thoughts?
>
>  >  Cheers,
>  >  -Lofton.  -----Original Message-----  From:
>  > 
> <mailto:public-webcgm-wg-request@w3.org>public-webcgm-wg-request@w3.org 
> [mailto:public-webcgm-wg-request@w3.org]
>  >  On Behalf Of Lofton Henderson  Sent: Montag, 20. Oktober 2008 12:06  To:
>  >  Bezaire, Benoit; WebCGM WG  Subject: RE: CGZ files
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >
>
>  >  Benoit,
>
>  >  Oops, my message crossed with yours.
>
>  >  In short:  I agree that this is a conformance requirement for 
> WebCGM  2.1,
>  >  specifically that 2.1 viewers must handle gzip-compressed 
> 2.1  instances,
>  >  and that valid 2.1 instances included plain Binary Encoding  as well as
>  >  gzip-compressedinstances of binary-encoded 2.1 metafiles.
>
>  >  Long analysis:  see my other just-sent message.
>
>  >  -Lofton.
>
>  >  At 11:23 AM 10/20/2008 -0400, Bezaire, Benoit wrote:
>
>  >  >I see your point, however...  >>We have customers using WebCGM 1.0
>  >  "compliant" tools (IsoDraw/IsoView  >v6 for example). Now, these 
> customers
>  >  could get a WebCGM 1.0 .cgz and  >those "compliant" applications would
>  >  reject them. That's not very  >user-friendly.  >>Maybe it's better to do
>  >  this as a WebCGM 2.1 feature.  >>Benoit.  >>-----Original Message-----
>  >  >From: Don L. [mailto:dlarson@cgmlarson.com]  >Sent: Friday, October 17,
>  >  2008 6:59 PM  >To: Bezaire, Benoit  >Cc: WebCGM WG  >Subject: re: CGZ
>  >  files  >>Benoit  >>> Hi  All,  >>>> I find the draft  underspecified 
> about
>  >  compressed CGM files. More  >>specifically, we would like to  know what
>  >  kind of CGM files may be  >>compressed?  >>>> Version1 to  4?  >> Can I
>  >  compress a  WebCGM 1.0 CGM file for example?  >>>> Is this a WebCGM 2.1
>  >  conformance feature for viewer and authoring  >tools?  >> Or is this 
> a new
>  >  WebCGM  2.1 (and only 2.1) 'encoding scheme' ...  >>for  >>>lack  of a
>  >  better  word?  >>I think 'encoding scheme' is a better characterization.
>  >  The text for  >this feature in the webcgm 2.1 spec was extracted from 
> the
>  >  SVG spec.  >>My thinking is that this is a viewer conformance issue 
> and a
>  >  WebCGM 2.1  >viewer should be able to open a file with a .cgz extension
>  >  and know  >that it needa to decode this file according to the gzip spec.
>  >  with the  >assumption that results will be a file that conforms to the
>  >  WebCGM  >profile (any version e.g.  >1.0 , 2.x).  >>Don.  >>> 
> Thanks.  >>
>  >  Benoit.
>  >
>
>  >
>