James Carlson wrote:
> Bart Smaalders writes:
>> Peter Tribble wrote:
>>
>>> Upgrading means testing and requalifying all my applications. At
>>> least under Solaris, patching never has. Under the current patching
>>> scheme, it's possible to get a fix for an individual bug; upgrading the
>>> whole package involves many more changes and significantly
>>> increasing the risk to the end user.
>>>
>> This imparts a much greater weight to the mechanism of software
>> change than is appropriate.  A patch can change every portion of
>> a package, or only a misspelling in a comment in an include file.
>> There is no semantic difference between a patch and upgrade; both
>> apply change to the system.  Whether or not you need to requal all
>> your applications depends on your sense of paranoia and consequences
>> of things going wrong.
> 
> What it *really* ought to depend on is our advertised content.
> 
> In other words, if we say that we're including only trivial changes
> that don't affect the way applications run, as we would usually do for
> patches, then it's as safe as it was before.  If we say that we're
> including life-altering breakage, then look out.
> 
> This comes back to release types (Major, Minor, Micro) and the
> labelling of what we ship.  I don't think that the project proposal
> needs to address that off the bat, but I do think there needs to be a
> clear and explicit way to know exactly what it is one is installing --
> preferably before it's installed.
> 
> Historically, people have aliased "patch" to mean "least level of
> change" and "Solaris Minor" to mean "applications need to be
> requalified because we don't quite trust that binary guarantee."
> That's only one implementation of the scheme.
> 
> Having "X.Y.Z" release numbers is one way to handle the problem, but
> there might be others.  I don't much care how it gets done, so long as
> the documentation of the content is clear and understandable.
> 

We've been talking about exactly such versioning of packages... this
will become clearer when we get an approved project and the docs and 
source become available.

-- 
Bart Smaalders                  Solaris Kernel Performance
barts at cyber.eng.sun.com              http://blogs.sun.com/barts

Reply via email to