PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Wyllys Ingersoll 

Somehow this case got created but the one-pager was never sent out.  Attached 
is the
proposal.  My apologies to the project team for the snafu.

-Wyllys



-- next part --
An embedded and charset-unspecified text was scrubbed...
Name: perl-crypt-onepager.txt
URL: 
http://mail.opensolaris.org/pipermail/opensolaris-arc/attachments/20100324/1863c5b6/attachment.txt

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Darren J Moffat 
I see no major issues with this case.

However the packaging will need to be different.

Probably one of these:

pkg:/library/security/openssl/perl/{aes,bignum,...}
pkg:/library/perl-5/openssl/{aes,bignum,...}

I think the second one is more likely as these are perl bindings for 
openssl.

-- 
Darren J Moffat

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Hai-May Chao 
Darren J Moffat wrote:
 I see no major issues with this case.

 However the packaging will need to be different.

 Probably one of these:

 pkg:/library/security/openssl/perl/{aes,bignum,...}
 pkg:/library/perl-5/openssl/{aes,bignum,...}

 I think the second one is more likely as these are perl bindings for 
 openssl.


I'm still using SVR4 packaging scheme.  I touched base with Norm,
and he suggested for me to use SVR4 definitions for now, until SFW
moves to IPS.  Depending on the timing, I'll either change it before my
putback or it will be changed to IPS with the rest of the gate.

Thanks,
Hai-May

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Garrett D'Amore 
On 03/24/10 11:07 AM, Hai-May Chao wrote:
 Darren J Moffat wrote:
 I see no major issues with this case.

 However the packaging will need to be different.

 Probably one of these:

 pkg:/library/security/openssl/perl/{aes,bignum,...}
 pkg:/library/perl-5/openssl/{aes,bignum,...}

 I think the second one is more likely as these are perl bindings for 
 openssl.


 I'm still using SVR4 packaging scheme.  I touched base with Norm,
 and he suggested for me to use SVR4 definitions for now, until SFW
 moves to IPS.  Depending on the timing, I'll either change it before my
 putback or it will be changed to IPS with the rest of the gate.

 Thanks,
 Hai-May


That's fine with me, but you need to fix the Committed value for your 
package names -- clearly they are not Committed, since you know you'll 
be changing them soon.

 - Garrett

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Hai-May Chao 
Garrett D'Amore wrote:
 On 03/24/10 11:07 AM, Hai-May Chao wrote:
 Darren J Moffat wrote:
 I see no major issues with this case.

 However the packaging will need to be different.

 Probably one of these:

 pkg:/library/security/openssl/perl/{aes,bignum,...}
 pkg:/library/perl-5/openssl/{aes,bignum,...}

 I think the second one is more likely as these are perl bindings for 
 openssl.


 I'm still using SVR4 packaging scheme.  I touched base with Norm,
 and he suggested for me to use SVR4 definitions for now, until SFW
 moves to IPS.  Depending on the timing, I'll either change it before my
 putback or it will be changed to IPS with the rest of the gate.

 Thanks,
 Hai-May


 That's fine with me, but you need to fix the Committed value for 
 your package names -- clearly they are not Committed, since you know 
 you'll be changing them soon.

 - Garrett


I'll fix it.

Thanks,
Hai-May

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Sebastien Roy 
On 03/24/10 02:09 PM, Garrett D'Amore wrote:
 That's fine with me, but you need to fix the Committed value for your
 package names -- clearly they are not Committed, since you know you'll
 be changing them soon.

Package names should be no less than Committed, especially now that IPS 
can refer to a package using an older name.  Since package names are how 
package dependencies are built, and how software is bundled and 
installed, anything less than Committed doesn't make sense to me.

That said, since SFW packages are delivered to the system using the 
hierarchical IPS scheme (e.g. diagnostic/wireshark), I don't think that 
the fact that the SFW gate doesn't directly deliver IPS packages is a 
valid reason for this case not to export a proper IPS package name, nor 
for the package name to be anything less than Committed.

-Seb

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Hai-May Chao 
Sebastien Roy wrote:
 On 03/24/10 02:09 PM, Garrett D'Amore wrote:
 That's fine with me, but you need to fix the Committed value for your
 package names -- clearly they are not Committed, since you know you'll
 be changing them soon.

 Package names should be no less than Committed, especially now that 
 IPS can refer to a package using an older name.  Since package names 
 are how package dependencies are built, and how software is bundled 
 and installed, anything less than Committed doesn't make sense to me.

 That said, since SFW packages are delivered to the system using the 
 hierarchical IPS scheme (e.g. diagnostic/wireshark), I don't think 
 that the fact that the SFW gate doesn't directly deliver IPS packages 
 is a valid reason for this case not to export a proper IPS package 
 name, nor for the package name to be anything less than Committed.

 -Seb

I'll update this case to reflect an IPS scheme (I'll check with Norm 
with regard
to the names of IPS packages).   The package names will be Committed.

Thanks,
Hai-May

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Nicolas Williams 
On Wed, Mar 24, 2010 at 02:22:56PM -0400, Sebastien Roy wrote:
 On 03/24/10 02:09 PM, Garrett D'Amore wrote:
 That's fine with me, but you need to fix the Committed value for your
 package names -- clearly they are not Committed, since you know you'll
 be changing them soon.
 
 Package names should be no less than Committed, especially now that
 IPS can refer to a package using an older name.  Since package names
 are how package dependencies are built, and how software is bundled
 and installed, anything less than Committed doesn't make sense to
 me.

I thought IPS supports pkg renaming, so why should pkg names be
Committed?

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Garrett D'Amore 
On 03/24/10 11:56 AM, Nicolas Williams wrote:
 On Wed, Mar 24, 2010 at 02:22:56PM -0400, Sebastien Roy wrote:

 On 03/24/10 02:09 PM, Garrett D'Amore wrote:
  
 That's fine with me, but you need to fix the Committed value for your
 package names -- clearly they are not Committed, since you know you'll
 be changing them soon.

 Package names should be no less than Committed, especially now that
 IPS can refer to a package using an older name.  Since package names
 are how package dependencies are built, and how software is bundled
 and installed, anything less than Committed doesn't make sense to
 me.
  
 I thought IPS supports pkg renaming, so why should pkg names be
 Committed?


Indeed.  There are more package renames and breakups in the works too -- 
the big change to hierarchical names was just the first round.

Furthermore, I seem to recall that IPS packaging in Open Solaris has 
automatic dependency tracking, which isn't so dependent on package names 
but on package manifests.

Btw, has anyone altered the docs folk that all those Availability 
attributes in the man pages are now obsolete?

 - Garrett

PSARC 2010/084 Perl Crypt bindings for OpenSSL

2010-03-24 Thread Sebastien Roy 
On 03/24/10 03:15 PM, Garrett D'Amore wrote:
 On 03/24/10 11:56 AM, Nicolas Williams wrote:
 On Wed, Mar 24, 2010 at 02:22:56PM -0400, Sebastien Roy wrote:
 On 03/24/10 02:09 PM, Garrett D'Amore wrote:
 That's fine with me, but you need to fix the Committed value for your
 package names -- clearly they are not Committed, since you know you'll
 be changing them soon.
 Package names should be no less than Committed, especially now that
 IPS can refer to a package using an older name. Since package names
 are how package dependencies are built, and how software is bundled
 and installed, anything less than Committed doesn't make sense to
 me.
 I thought IPS supports pkg renaming, so why should pkg names be
 Committed?

 Indeed. There are more package renames and breakups in the works too --
 the big change to hierarchical names was just the first round.

 Furthermore, I seem to recall that IPS packaging in Open Solaris has
 automatic dependency tracking, which isn't so dependent on package names
 but on package manifests.

At the risk of outing 2010/067 before the materials are made open, that 
case suggests that software use package versions to track feature 
support for various components.  This is another place where package 
names are codified.  In any case, the existing ARC best practice trumps 
this discussion.

http://sac.sfbay/cgi-bin/bp.cgi?NAME=stability.bp

If we need to change the best practice, let's not do it here as part of 
this case.

 Btw, has anyone altered the docs folk that all those Availability
 attributes in the man pages are now obsolete?

What does this have to do with this case?

-Seb