>> Did anyone consider staying with SVR4 and just making the tools work 
>> properly.
>> i.e a green field site and using SVR4 build a complete set of tools from 
>> ground up.
>> Enda
>> 
>
>Based on all already published info I saw - no in reality. SVR4 is not
>broken itself but they think that some new packaging system will solve
>their ignoring of feature delivering magically somehow, ignoring the
>flexibility of SVR4.


There is unfortunately a history at Sun to throw away things and start
from scratch.  While the "from scratch" bits are sometimes better designed
the "we'll implement lost functionality in stage 2" never materializes.

The prime example of this, I believe, are the admin tools:

In Solaris 2.1 there was a reasonable set of tools which worked locally
on a system, including a graphical package installer which knew about
clusters, what was installed on the system and allowed you to simply add
back what was missing or remove what you didn't want.

When this was rewritten in a "networked based" admintool shell, apart
from the huge security issues this introduced, that functionality was
never restored.

With the removal of admintool even more functionality was gone and
both SMC and lockhart supposedly stepped up to the plaid but the
former sees no development "because it is being replaced" and the
latter is "only a framework" so it's completely useless.

Whether this is going to happen this time, I don't know, but I see
this as a real risk.  My starting position is therefor generally
"only rewrite from scratch if you can proof you cannot sufficiently
improve".

We have proven that install slowness is caused by bunzip2; that is the
low hanging fruit.

Casper


Reply via email to