Call for feedback on a Ports-collection change

[Garance A Drosihn]

At 9:47 PM -0800 1/8/04, Chris Pressey wrote:
>On Fri, 9 Jan 2004 00:07:53 -0500
>Garance A Drosihn <drosih at rpi.edu> wrote:
>
>>  Inside the proposed pkg-data file, the patches can have
>>  whatever names you want.
>
>Names that you will rarely, if ever, see in a unified diff
>between two pkg-data files - I think that's the most common
>objection so far.

Well, we could fix 'diff' for that...  :-)

>  > Not only that, but you're killing performance when doing
>>  operations on the entire ports tree (an operation such as
>>  'cvsup').  The amount of time to find-and-read ten small
>>  files is going to be much more than to find-and-read one
>>  larger file (particularly if that entire larger file can
>>  still fit in a single block on the disk).
>
>So fix cvsup :)

It isn't an issue with CVSup.  It's the reality of disk I/O.

There are already people talking about how we will soon
have hard disks which are huge and fast, but which you will
*never* be able to fill up if you're writing billions of
little random-access files.  Seek-times and look-up times
will completely overwhelm any actual data-transfer times.

I'm not saying *that* is a valid justification for doing my
project, but I'll try to subliminally get you to think that
it is significant without me ever actually saying it is... :-)

>  > Let me just say that I have some long-term ideas which are
>  > a bit more ambitious, but this idea is a doable first-step
>  > towards those ideas.
>
>Ah, well that's a slightly different goal than "Just reduce
>the inode count" isn't it?

Well, the inode-count is (IMO) the main justification of *this*
project.  I didn't want to try to sell this project based on
some follow-on project, when I have no idea that I'll ever get
to any follow-on projects...

>Sorry, I don't mean to sound snarky,

No problem.  I didn't want to write up a huge message with
all the details until I had some feedback on a simplistic
overview of the idea.

>but when you stated that goal I took it on face value...
>and now you go and change it :)
>
>But actually, I don't see how bundling everything up into a
>single (or a couple of) pkg-data file(s) leaves the ports
>tree in a more flexible state, either...

Well, because you can then add more fields to the new file,
without adding more files to the overall ports collection...

-- 
Garance Alistair Drosehn            =   gad at gilead.netel.rpi.edu
Senior Systems Programmer           or  gad at freebsd.org
Rensselaer Polytechnic Institute    or  drosih at rpi.edu