[lsb-discuss] Branching the "Needed network interfaces" discussion

[Robert Schweikert]

<snip>
So what do we do with commonly-used interfaces that
have not yet been fully replaced in code in the field
by newer, more capable, and "preferred" interfaces?  

I think it's pretty clear from this example what the
LSB selection philosophy has been in the past, where 
we were aware of a conflict situation like this. We've
been informally advised to follow this approach by 
upstreams, but it does leave situations where code may
have to be reworked before becoming LSB conforming,
or as in another case Robert has brought up, may cause
problems for code obtained from another vendor that
may not be available to modify.
<snip>

There is no escaping the install base and usage of old interfaces, there
are plenty of examples of this problem. Maybe a fourth category of
interfaces is required. Today in my mind there are 3 interface
categories:

1.) supported (part of LSB)
2.) not supported
3.) deprecated

When running appchk application vendors should probably get PASS, FAIL,
and WARNING messages respectively for these categories. WARNING messages
should be followed by a time frame for example

WARNING: Symbol X will be removed from the LSB in version 6.2

this way when 6.2 rolls around vendors (depending on the length of the
deprecation policy) had enough time to clean up their act, yes they will
complain but that will always happen even if the deprecation warning
shows up for 10 years in a row. 

Anyway, back to the fourth category. Interfaces discouraged for use by
the LSB, the IPV4 network interfaces (gethostbyname_r etc.) being
discussed could fall into this category. 

INFORMATION: Symbol gethostbyname_r should not be used and is superseded
by getnameinfo

However, this also means there has to be yet another policy about what
to do with discouraged interfaces. Obviously as deprecated interfaces
these cannot be carried around forever and this is where all the
discussions merge. 

Another option of course is that having a long deprecation period may be
sufficient and the fourth category is not needed.

Maybe one or more of the Tuesday calls can be devoted to this topic.

Thanks,
Robert
-- 
Robert Schweikert                   MAY THE SOURCE BE WITH YOU
(Robert.Schweikert at abaqus.com)                 LINUX
ABAQUS Inc.
Phone : 401-276-7190
FAX : 401-276-4408