Red Hat Diaries/0081

Hacker

From: Rixster
To: Scot Hacker
Subject: Re: iPhoto, iTunes Falling Down on Library Size
Date: Mon, 20 Jan 2003 14:41:59 +0000


Re: iPhoto, iTunes Falling Down on Library Size
http://www.oreillynet.com/cs/weblog/view/wlg/2558

I can't say I'm surprised. Over Thanksgiving last year I made a few discoveries in the Cocoa API - namely that they're peremptory and not built for adaptable, easy, flexible use. I never thought I'd be caught singing the praises of Microsoft software designers, but I was.

The object of my concern was the Cocoa NSTableView, which is used in listings such as you see in iTunes. This is a monster, with little or no built-in support for application/control integration. It just sits there. And the reason no one has objected until now is that no one has tried to use it in serious high-scale business applications.

We did a rudimentary test of the NSTableView and the concomitant Cocoa API: We took code from an 'accepted guru' (Aaron Hillegass) and constructed a table with 40,000 entries (rows) with seven (7) columns each. (Loading in this data was not difficult, because the test program created 'random' data on the way in. Actually reading in data from disk and feeding the NSTableView would have made the computer grind to a halt.)

Once the data was resident, the trick became to remove all but one of the entries (rows).

Naturally, we had a comparable app ready to benchmark on the Windows platform.

Machines
--------
Windows: Running Windows 98SE (16-bit OS) at about 150MHz (yes a very slow machine).

Apple:   A 'TiBook' running OS X (32-bit OS) at 550MHz.

Results
-------
Deleting all but one of the entries (rows) in the Windows test program took just over one minute real time.

Deleting all but one of the entries (rows) in the Apple took over seven (7) minutes.

* * * * *

Naturally, I was worried, and after considering the data for several days, decided to publish my findings at the Cocoa-dev mailing list. That was a big mistake.

You'd think the Apple developers would be more interested in getting their platform up to par. You'd think they were more interested in finding out why where and when their platform was not 'perfect', and in which ways they could improve its ability to scale.

But then again, you might not think that at all. You might already know better.

Click here »