RFC: A new git repo for updates to the gABI

[H.J. Lu]

Hi,

The gABI:

http://www.sco.com/developers/gabi/

is virtually unmaintained for many years. But we still need to
update gABI from time to time. I propose to create a new git
repo to document updates to the gABI. We can use Linux gABI:

https://gitlab.com/x86-psABIs/Linux-ABI

as a template by stripping out Linux specific things. We can
put it on gitlab.com.

--
H.J.

[Cary Coutant]

It's gonna be more than stripping out the Linux-specific things -- you
need to add in all the generic ELF things which are not in that
document.

I'm working on a new from-scratch version of the gABI, and was
planning on hosting it on github eventually. I know it's been a while,
but I do plan to get this done. Please be patient for a while longer.

I did ask Xinuos for the rights to use the existing text, but that was
unsuccessful, so I've gone with plan B.

-cary

> --
> You received this message because you are subscribed to the Google Groups "Generic System V Application Binary Interface" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to generic-abi...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/generic-abi/CAMe9rOpwXFrFVuuApGTjWtWL8V5NpUG%2BmpXyzq%2BuE6QrKOWnoA%40mail.gmail.com.

[H.J. Lu]

On Mon, Mar 23, 2020 at 4:07 PM Cary Coutant <ccou...@gmail.com> wrote:
>
> It's gonna be more than stripping out the Linux-specific things -- you
> need to add in all the generic ELF things which are not in that
> document.
>
> I'm working on a new from-scratch version of the gABI, and was
> planning on hosting it on github eventually. I know it's been a while,
> but I do plan to get this done. Please be patient for a while longer.

I am very glad you are on it. It won't hurt to wait a little while longer.

> I did ask Xinuos for the rights to use the existing text, but that was
> unsuccessful, so I've gone with plan B.
>

Thanks.

--
H.J.

[Michael Matz]

Hello,

On Mon, 23 Mar 2020, Cary Coutant wrote:

> It's gonna be more than stripping out the Linux-specific things -- you
> need to add in all the generic ELF things which are not in that
> document.
>
> I'm working on a new from-scratch version of the gABI, and was
> planning on hosting it on github eventually. I know it's been a while,
> but I do plan to get this done.

\o/ Hoorah! :-) (If you want any help (proof reading, writing boring
texts, whatever), count me in)


Ciao,
Michael.

[Carlos O'Donell]

Thank you Cary! I am also happy to proof read and write boring texts.

I spent forever trying to understand ELF dynamic string token handling
given the language of the original text.

--
Cheers,
Carlos.

[Cary Coutant]

Michael & Carlos, I will definitely take you up on your offer! Thanks!

-cary

> --
> You received this message because you are subscribed to the Google Groups "Generic System V Application Binary Interface" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to generic-abi...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/generic-abi/bf6fd47d-4313-9b3d-9c13-5f1df90268ae%40redhat.com.

[Ali Bahrami]

I'll also volunteer, if you need the extra eyes.
I bring a non-GNU perspective that might be helpful.

In terms of rewriting, IANAL, but I wonder if we could
just start with a copy of the existing gABI, that gives
the original authors copyright credit. A concern I have
about a rewrite is that of drift relative to the original.
Starting with the existing text would help prevent that,
and would be a lot easier as well.

- Ali

[Ryan S. Arnold]

IANAL. When I was at IBM and working on the (then) new PowerPC 32-bit ELF ABI supplement, we (the now defunct power.org) ran into this issue.  We wanted to start from the pre-existing SysV ABI PowerPC Processor Supplement which was originally a joint document copyright IBM and SUN Microsystems.

Unfortunately, this was during the acquisition of SUN Microsystems by Oracle and our attempts to contact them came to naught due to the chaos of converging companies.

More troubling, there were sections of the psABI that "shared text" with the gabi (System V Application Binary Interface). Considering SCO's litigious nature at the time it was considered very risky to attempt to derive from the existing psABI supplement at all.

In an abundance of caution, we found no alternative other than to write an entirely new document from scratch (released under the GFDL 1.3 to prevent this from happening again and to provide clean provenance going forward)

We were advised that the only thing that could be reused were statements of fact, i.e., tabular data (lists of fields, etc)

In the end we cross checked against the original document for consistency.

Given SCO/Xinuos's litigious past, deriving from something clearly marked "All rights reserved" that expressly states "Copyright infringement is a serious matter under the United States and foreign Copyright Laws" without a copyright grant seems unwise.

For reference, here are the GFDL sources for the Power Architecture 32-bit Application Binary Interface Supplement:

https://github.com/ryanarn/powerabi

Regards,

--
You received this message because you are subscribed to the Google Groups "Generic System V Application Binary Interface" group.
To unsubscribe from this group and stop receiving emails from it, send an email to generic-abi...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/generic-abi/ebfe7c1a-41d9-1ae2-fcd4-402718993819%40Oracle.COM.

[Cary Coutant]

> IANAL. When I was at IBM and working on the (then) new PowerPC 32-bit ELF ABI supplement, we (the now defunct power.org) ran into this issue. We wanted to start from the pre-existing SysV ABI PowerPC Processor Supplement which was originally a joint document copyright IBM and SUN Microsystems.
>
> Unfortunately, this was during the acquisition of SUN Microsystems by Oracle and our attempts to contact them came to naught due to the chaos of converging companies.
>
> More troubling, there were sections of the psABI that "shared text" with the gabi (System V Application Binary Interface). Considering SCO's litigious nature at the time it was considered very risky to attempt to derive from the existing psABI supplement at all.
>
> In an abundance of caution, we found no alternative other than to write an entirely new document from scratch (released under the GFDL 1.3 to prevent this from happening again and to provide clean provenance going forward)
>
> We were advised that the only thing that could be reused were statements of fact, i.e., tabular data (lists of fields, etc)
>
> In the end we cross checked against the original document for consistency.
>
> Given SCO/Xinuos's litigious past, deriving from something clearly marked "All rights reserved" that expressly states "Copyright infringement is a serious matter under the United States and foreign Copyright Laws" without a copyright grant seems unwise.

Yes, SCO had a reputation for being litigious, but I think Xinuos has
done their best to distance themselves from that.

Regardless of SCO's reputation, your lawyers' advice was sound (as one
would expect at IBM!). Corporate legal departments do get accused of
being overly cautious at times, but I don't think this was a case of
being overly cautious at all. Those materials were copyrighted, and
even a non-litigious company will want to defend its IP (a public
company has a duty to its shareholders to do so).

In fact, I went through the very same process in documenting the HP-UX
ABI: I started from scratch and cross-checked. Unfortunately, all that
material I wrote remained under an HP copyright. That's the material
that Suprateeka is now working to release under a more permissive
license.

I'll have more to say on the future of the gABI documentation in a new post.

-cary