Good points from everybody. Adding a few tidbits of my own. Just to clarify
I'm no SVR4 lover having had worked on the messy codebase during my stint
at SUN. It had good ideas but the age of the code was showing.

1) In essence packaging and package distribution is a solved problem. The
big
   question of why re-write a new thing when several excellent open source
   alternatives existed was never answered. No analysis was provided.
Nexenta
   demonstrated the technical feasibility using the Apt toolchain
successfully
   though there were licensing issues. There are other more friendly
upstreams
   with better licensing as well.

2) Assumptions were made, design decisions taken without involving people
   who had experience with distributions. Mistakes were made silent U-turns
   taken and so on.

3) I was shown the door when I started asking questions:

http://mail.opensolaris.org/pipermail/pkg-discuss/2008-September/006509.html

http://mail.opensolaris.org/pipermail/pkg-discuss/2008-September/thread.html
   (Search for subject "Observations on IPS")
   If that was how community was dealt with imagine how it was dealt with
inside
   the company. In fact I got the guts to ask questions only after leaving
SUN.

4) As many have pointed out, no scripting is desirable but does not require
    re-inventing a new package management solution from scratch. The action
    mechanism IMHO was poorly thought out requiring SMF services to handle
    some postinstall mechanisms (like Gnome related stuff). Using the
system's
    service manager to handle package requirements is so broken.

5) The implementation grew and grew in complexity. There was ample scope
    creep. Which package manager in the world had to develop it's own custom
    HTTP server?

6) Cross-platform support was provided including support for Windows for
    managing the Java app installation on Windows. Isn't this another scope
    creep for a Unix packaging solution?

7) Performance was bad especially if one is remote from the Pkg server. For
    example Installing a 1MB package from machine in Bangalore from a Pkg
    server in California used to take long. Especially if one is trying to
install after
    a gap of say a month. It would spin in plan creation for minutes.
    Downloading an ISO image of the OpenSolaris livecd from the same
location
    showed better bandwidth utilization!

8) I'll make a controversial comment here. Which technology organization
takes
    some of it's smartest core OS engineers and puts them on a multi-year
project
    to develop a new all-encompassing multi-platform package management
    solution?
    With a practical approach and with people having experience in these
matters
    it probably takes 6 months to a year to roll out something that solves
the
    problems and is stable.

    As a side note I had worked out a full-featured package manager for
SVR4 in
    about 3 months of part-time effort. This was for BeleniX. Not that I
liked to
    remain on SVR4 but I had little option at that time.
    See:
    https://moinakg.wordpress.com/2008/11/22/the-belenix-package-manager/
    Point #9 "Recursive Uninstall" was actually discussed by the pkg-discuss
    folks after I made that post.

Regards,
Moinak.
-- 

P.S.: Running for cover. I might have stirred an old hornet's nest.
================================
http://moinakg.wordpress.com/
http://moinakg.github.com/pcompress/



-------------------------------------------
illumos-discuss
Archives: https://www.listbox.com/member/archive/182180/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4
Powered by Listbox: http://www.listbox.com

Reply via email to