On 03/19/10 04:39 PM, Alan Coopersmith wrote:
> [Removed the case id, since this is off-topic for the case which isn't 
> currently
>   on the table for discussion anyway.]
>    

And I removed PSARC. :-)

> Garrett D'Amore wrote:
>    
>> I'm also of the opinion that it is a mistake to sacrifice familiarity
>> for our paying Solaris 10 customers in favor of familiarity for people
>> coming from Linux.
>>      
> But clearly all our paying Solaris 10 customers already have dotfiles to
> set $PATH, given how useless the default Solaris 10 $PATH is.
>    

Not when they create a new user account.  Yes, its easy to override.  
But should they have to do that?

(Also, its not entirely unheard of for a customer to use the supplied 
default PATH.)

> They get familiarity by continuing to use those - the default PATH with
> /usr/gnu/bin first only affects those setting up new accounts who don't
> have existing Solaris .dotfiles, which seems like a very reasonable
> compromise.
>    

The problem is that creation of a new user account does not necessarily 
mean a new *person*.  And, furthermore, the change in default means that 
sites are going to have to deal with the differences -- instructions and 
scripts provided at sites for new users may assume a different default.

How many times have you had to debug a user problem that amounted to a 
different (and unexpected) element in $PATH?  (Think how much pain 
/usr/ucb caused over the years).   We should be able to provide a 
reasonable that default that makes most folks happy.  /usr/gnu is *not* 
that default, for a variety of reasons.

Pretending that nobody is going to be negatively impacted by the 
promotion of /usr/gnu as the new default is a mistake.

> Also rememeber the PATH default is set only in text files which are trivially
> editable by users with experience from previous Solaris releases - it's not
> baked into the kernel.
>    

True enough, but remember that sites are going to have to deal with 
documented procedures, etc.

In any case, I'm not advocating that we confine ourselves to only 
supporting the legacy commands -- I know that much of /usr/bin was 
bit-rotten.  I am however advocating that /usr/gnu is *not* the 
end-all-and-be-all of the story; and that by capitulating to GNU we also 
limit the kinds of innovation we can do in the future.  (What about 
adding new kinds of file system objects?  Can we get the GNU users to 
adopt extended attributes, etc?  If we supply our own versions -- using 
the ksh93 code as a base -- we don't have that problem.)

>    
>> Which group do you think contributes more towards
>> the $$ that pay our salaries?
>>      
> Sounds like an invalid question for PSARC-ext&  shell-discuss at 
> opensolaris.org.
>    

You're right -- its out of scope.  Still, its something I wish folks 
would consider -- the folks who actually buy large amounts of 
Sun^WOracle equipment in my opinion deserve at least as much 
consideration in our compatibility concerns as the theoretical Linux 
immigrants.

     - Garrett

Reply via email to