Saturday, January 23, 2010

Living with technology regressions in the post-performance era

I'm slowly getting used to living with the post-Moore's Law era of technology regressions.

It's taken a long time to get over my early computing experience. Switching from an 8086 to an 80386 in the DOS era was pretty much pure progress. The transition to OS/2 then to Windows 95 involved many regressions, but I imagined that was a one time anomaly. Win 98 to NT to 2000 to XP was pretty much all improvement (I skipped ME of course).

Same story in my early Mac days. Things just got better - until MacOS 7 ran into TCP/IP. That was a train wreck, but it did get sorted out. When I returned to the Mac I was using OS X 10.1 (or 10.2?) and that was good too.

Ok, so some great software died without replacement. I should have adjusted my expectations. What can I say? I'm a geek. CPUs kept getting faster. It helped me overlook a lot of things.

Alas, the decrepit state of the Wikipedia entry on Moore's Law speaks volumes. We may get more transistors, but clearly we're not getting more performance. We're in the post-performance era.

In our new era some things get better, some things get worse. Personal computing is middle-aged. Progress is uneven.

I'm in the midst of one of those tech churn transitions now with my backup systems.

I'm not paranoid about backups, because the universe really is trying to destroy my data. I'm just realistic.

Realism means I've long had fully automated rotating off-site backups, and, as backup software quality has regressed, I've moved to having two completely distinct automated backup systems. (If two distinct systems each have a 90% reliability rate, then the probability one will work is 1-(0.1*0.1) or 99.9%. It's almost impossible to equal that reliability from a single affordable product.)

So I probably still have a reliable backup system, but it's more work to maintain than my old system. In some ways it's also less flexible, in particular my laptop backups are less reliable. I'm having to adjust my workflow to the new environment, and that means some functional regressions.

Middle-aged post-Moore computing means living with regressions. The trick is realizing when a true functional regression has occurred, and then being able to say good-bye to the better for the sustainable.

Update 1/25/10: I just found this 2006 tech post of mine complaining about the backup market. It's been a bad few years for backup. I should also highlight a comment Andrew W made (see below):
On Windows there's Home Server, which is about as carefree a centralized/networked backup solution as you can get.
The Apple equivalent to Home Server is Time Capsule. I would like to see Apple do a more complete backup/media server/file server home server solution.

2 comments:

Andrew W said...

Tech *should* be allowed to die.

Retrospect is a great example. It's horribly antiquated software that, despite sold to home users, it never adequately targeted them. It is extremely opaque and essentially took a "home IT staff" to keep operational.

That software was left behind because the market has evolved. Unlike 7 years ago, end-user backup is now a reality.

On Windows there's Home Server, which is about as carefree a centralized/networked backup solution as you can get. There's also plugins for simple backups to S3. And HP even has extensions for Macs. All for, what, $400?

On Mac there's Time Machine and Super Duper -- I use both, fairly painlessly. I read your problems with TM, but the bottom line is that TM creates a "frictionless" copy of your files, over time, on a secondary drive. Maybe it's not perfect, but *any* backup is better than what most users were running 7 years ago. If, in some cases, it takes a little bit of "expertise" to get your data out then restores are hopefully a rare case.

The point is that "end-user backup" is much more of a reality then it ever was with Retrospect and other products of that era. In most cases an unsophisticated would give up before getting it working at all.

And hence that product has been sent to the Island of Unwanted Software where it belongs.

I'd argue that the worst problem with your archive is that you've locked your photos into iPhoto's proprietary database and bizarre file structure. You'll have a hard time getting your photos out w/out iPhoto. And if the database file(s) are somehow corrupted then you could lose *all* albums & metadata. It'd be much safer to use a simpler file/folder/embedded-tags archive for such important content. Lock-in is especially painful when it's all-or-nothing.

JGF said...

I was using Retrospect Pro on Windows. The version I'm on was pretty good - much better than Retrospect 6 or 8 for OS X. I didn't mind the complexity of Retrospect since i knew it pretty well. Most of all it had actually saved me several times, so I respected it.

I've had many more bugs with Time Machine than with Retrospect/Win. So losing it is a regression for me. (Retrospect 8 appears to have been abandoned by EMC).

I made my iPhoto commitment with full knowledge of Apple's data lock habits. It's popular enough, and scriptable enough, that I figured there's be solutions to get much of my data out if Apple made continuing with it untenable.

These days though I might go with Lightroom or Picasa instead.