Emails like the following, are what add to the confusion and 
misunderstanding about IPS, postinstall scripts, SMF-foo,etc, etc :-)



Danek Duvall wrote:
> ...
> 
> Yes, it's code: it can do *anything*.  Nothing whatsoever can prevent that.
> ....
> 
> You just won't be able to do it in precisely the way you
> used to, and you'll be constrained to doing it in only one well-defined
> context, which should simplify life for the packager trying to construct a
> well-behaving package.
> 

So, in one paragraph, you say "it can do anything".
in the next paragraph, you say "you'll be constrained". (without any details 
to "the folks out here" as to what kinds of "constraints" are involved here)


Can you see my concern now?

This is not the first time this sort of apparent contradiction has been 
written on the mailing list on this subject. Which is why it has been 
unclear. and why it is important to write it up in a standalone document 
describing all this stuff!

When faced with reading "blog" info, etc. that is potentially a year old or 
whatever, vs reading something on this mailing list, that is presumed 
"current"...
  with no official "this is IPS and how it works" document, people tend to 
go with what they read here as the more accurate.
Which unfortunately, may lead to more confusion.





PS:
>   If it's not a reversible operation, then
>> that fault lies with the packager.

There is not neccessarily a "fault" of the packager. Some things are simply 
unavoidable. Such as the db data upgrade example I gave. Some non-reversible 
things must happen, to use a product at all.
If the packaging system claims by default, "this install is completely 
reversible!", and gives no mechanism to the packager (auto-detected, or 
manually flaged, either way), to declare, "nope, this is not default; this 
install is NOT reversible", then that is a lacking in the packaging /system/.

Even if you argue, "well, data migration should be handled in a separate 
script, and run separately by the site admin";
That action has just rendered the pkg install, non-reversible.
Yet the IPS pkg infrastructure will presumably still claim, "sure, you can 
reverse this install!"

I'd say avoiding false "this is reversible" claims, is an argument *in 
favour of* integrating data version upgrade scripts, with the packaging 
system. It should lead to more accurate reporting of whether non-reversible 
actions have been taken or not.



_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to