The BFC Computing Weblog

WWDC prediction - hey, all the kids are doing it.

Solaris, Open Solaris.

Under Mac OS X. Briefly:

• xnu/mach is really slow. Linux blows away OSX when threads are concerned. Solaris doesn’t have this problem
• ZFS: Sun already announced that OSX will default to ZFS as its filesystem.
  • This is a Solaris technology.
  • Tested code is essential.
  • Steve Jobs doesn’t let anybody else give away the biggies. Just ZFS is therefore a minor part of the announcement.
• DTrace: Looking on the ADC site, there are WWDC sessions on DTrace, another Solaris techonology
• Nexenta is Ubuntu on OpenSolaris - proof that you can skin a kernel
• OpenSolaris to go GPL3. Jonathan Schwartz has said as much, assuming he likes the final license
• Here is a picture from the WWDC lobby. A Sun is behind OSX. Game, set, match.

I just notice over at Businessweek, there’s a big shakeup at Palm today. They sold a quarter of the company to an investment group, kicked out the CEO, and Jon Rubenstein, former hardware VP at Apple, is taking over and is going to be working with Jeff Hawkins on their new products.

For those of you who don’t assiduously watch this space, that new product is a linux-based smartphone to launch at the end of this year.

Apple went too early with the iPhone, wanted to do something new with cell phones (good), got turned down by Verizon (expected) and got stuck on Cingular (bad). Now, you have to be a Cingular customer to own an iPhone and they’re in that deal for five years. This will be the first category of Apple products I won’t own. There ain’t no service here on AT&T and if I lived five miles to the right they couldn’t even write me a contract.

I will, however, buy Palm’s answer to the iPhone, assuming it’s no worse than my Treo (that would be hard). I think they learned some of the ‘obvious’ clues from Apple, the networks all get that they missed out on the iPhone, and they’ll work with Palm. So the iPhone will be available on AT&T and the Palm Treo replacement (Quatro?) will be on Verizon, Sprint, Nextel, MVNO’s, Skype, SIP, etc.

Let me tell you about this device called a Newton…

My nice (hardware-wise) Canon LIDE-30 scanner has no Intel Mac drivers. Canon has decided they won’t support Intel Macs on this hardware because it’s too old to write new drivers for, even though they sold it since the Intel transition and fully support it under Vista.

There’s no notice of this on their website, so I spent an hour or so trying to run the current version (with modification dates well before the Intel announcement) and the Installer craps out. Some people on VersonTracker claims it works fine, but over at Macintouch there are people who’ve heard direct from Canon say it ain’t so. Perhaps some folks are confused about the CPU on their Mac.

Anyway, I’m more than happy to dump Canon for their poor support policies, but I haven’t yet found a bus-powered flatbed scanner other than the LIDE series (recommendations in comments welcome).

I tried VueScan, which seems to work, but the demo mode peppers the scan with so many $ signs as a watermark that I couldn’t really evaluate the scan quality. So I’d have to pay to evaluate it and it’s certainly not open source. I don’t do enough scanning for that bargain to be worthwhile.

Then I remembered using the SANE scanner program under linux a few year ago, so I fired up VMWare Fusion on my Mac and ran my Fedora VM, and tried xsane. It worked great. I used sshfs to copy the files back to the Mac filesystem, and thought I had a decent solution.

But wait - I also have fink installed on my Mac, which has ports of lots of unix software. So, under X11, I ran:

fink install xsane
          

and it asked me if I wanted to install the dependencies (ya, I want Cheesy Poofs) and after ten minutes it was done compiling.

So I ran xsane and it said, “hey, you’ve got no scanner (try running as root)”. So I said, sudo xsane, and it said, “hey, this will melt your computer and cause flooding and plagues, but you’ve got LIDE-30”. Actually it just opened up a warning note and a stack of windows, but one of them said LIDE-30 in the menubar. If anybody knows how to access the USB device as a regular user, please leave a comment. This issue also prevents Scan Again from working, I think - try it if you don’t have this problem.

Now, there are two things you can do - scan in 16-bit and get the best image size - and have almost nothing be able to understand it, or scan in 8-bit and have everything understand it. I’m doing FAX’es here, not photos, so 8-bit is what I wanted. Change the bit depth to 8-bit:

Now, I want my stuff to work in Preview. Preview has trouble with all kinds of file formats that work under Linux tools, but for now we’ll adjust xsane to make a TIFF type that Preview can understand. Check the boxes to reduce to 8-bit and change the 8-bit image compression to ‘pack bits’.

Now, the xsane interface is a bit non-intuitive. In the preview window, grab the selection box to set the scan area. On the main screen, set the resolution next to the dotty-looking icon. It’s ‘200’ in this image:

Now, change to mode to ‘Save’, the color depth you need, the file type to ‘TIFF’, and click ‘Scan’ (you can preview first if you need to crop). Now go look for your out.tiff file in the directory where you launched xsane, and use it as you want.

Note: under the Linux version you’ll get a PDF save type and also have the ability to do a multiple-document batch. However, Preview strikes again, so you’ll have to post-process the pdf with:

pdf2ps inputfile.pdf - | ps2pdf - finalfilename.pdf
          

so Preview can understand it. If Adobe Reader is good enough you don’t have to convert it.

I barely ever send a fax. Ironically, I’m in an office building now with no analog phone lines available, so I’m setting up lots of VOIP stuff with Asterisk/Trixbox and the vendors want forms faxed to them. Ironic, since FAX is one of those things that VOIP vendors run away from scared when asked. I understand it’s possible with the ulaw codec and a sprinkling of luck, and I’m sure I’ll tackle it someday, but I needed to send some faxes, not tackle something.

I used to have an eFax account for years, until one day I didn’t. I did the usual round of searching, reading message boards, and examining TOS statements, and ultimately found TrustFax from Comodo Group. I used to use Comodo Group for my SSL needs until their prices skyrocketed. Besides the price, they were good to work with.

So, here’s the deal: for $30 upfront ($2.50 a month) you get 150 FAX pages, whichever direction you want. You get a toll-free number, a web interface, an e-mail interface (an API of sorts) and, they claim, no junk faxes.

I’m a bit skeptical about the last bit, but maybe because it’s a toll-free number that has something to do with it. I dunno. I’ll be happy if it’s true. At twenty cents a page, I’m going to complain if it’s not.

So, yeah, per-page this is expensive. But it’s their cheapest plan and run by a decent company, so if it works, 150 pages is more FAX than I’ve ever sent in my life, so it’ll be plenty.

And I just sent two faxes off without a hitch. OK, there was a hitch. My nice (hardware-wise) Canon LIDE-30 scanner has no Intel Mac drivers. Canon has decided they won’t support Intel Macs on this hardware because it’s too old to write new drivers for, even though they sold it since the Intel transition and fully support it under Vista. Anyway, xsane under linux (running under VMWare, even) saw it just fine, and I scanned it to PDF (run the pdf through pdf2ps inputfile.pdf - | ps2pdf - finalfilename.pdf to make Preview see it OK.)

When I’m in the mood to tackle something I’ll figure out FAX over VOIP - I’ve got a year to worry about it, though. Sometime during that year I’ll also find a new scanner - not made by Canon.

While descriptive of how things may look to the world as I move my gear into my new office, this blog entry is about an annoying error message on Mac OS X Tiger.

Cutting to the chase: You get this when Finder launches if it can’t get a route to the network server of the items in your

~/Library/Recent Servers
          

folder. Delete them, and you won’t have to hit Enter twenty times on login. Design note to Apple - don’t annoy the user when a backend task gets confused unless it’s critical.

Apple hyped Quartz Extreme, 3D acceleration on Mac OS X, quite a bit when Tiger was released. But they ship Mac OS X with 2D acceleration disabled. (Wha?)

Apparently on some hardware and with some software there are rendering glitches. On my Macbook Pro with all the software I use I haven’t seen any yet. And with it on, many applications, especially Firefox which does some ugly things with lots of 1x1px hidden windows, are several times faster than without the acceleration.

To turn it on temporarily, install the Developer Tools, launch QuartzDebug and enable it from the menu. You have to launch an app after it’s been enabled to get acceleration.

When you quit QuartzDebug it’ll be disabled again. To turn it on permanently (until you launch QuartzDebug again, anyway) execute this command:

sudo defaults write /Library/Preferences/com.apple.windowserver Quartz2DExtremeEnabled true
          

Once you’re satisfied that it works well, if you use an LCD display, try turning off BeamSync - a rendering delay built in to sync the screen refresh with the CRT electron beam sweep. You can test this in QuartzDebug or turn it off via the command line:

sudo defaults write /Library/Preferences/com.apple.windowserver Compositor -dict deferredUpdates 0
          

You can also edit these with PropertyListEditor, but you’ll have to give the admin group write access to it first:

sudo chmod 664 /Library/Preferences/com.apple.windowserver
          

Note, QuartzDebug resets this to 644 when you launch it.

One of the more interesting features of Mac OS X is its Spotlight - a system-wide metadata database (metadatabase?) that was obviously rushed out the door before it was completely baked.

For a couple weeks I haven’t been able to do full-text searches of my e-mail store - it just stopped working. I found in the Spotlight preferences, the hard drive was added to the ‘Privacy’ pane (not by me) and it couldn’t be removed. Looking at the actual files (/.Spotlight-V100/) the modification times were changing but the size of the database files never did. Nothing in the logs about it.

After several dead ends, I finally figured out that one just needs to remove all the Spotlight files and reboot. One nice advantage of the Spotlight architecture is this is non-destructive - everything will be re-built automatically. One disadvantage is everything has to be re-built automatically, so it’s resource intensive. Anyhow, the exercise distills down to:

sudo mdutil -i off /
          sudo mdutil -E /
          sudo rm -rf /.Spotlight-V100/
          

Then restart the computer. This article from The X Lab is the most comprehensive, accurate, and intelligible reference to Spotlight workings I found on the ‘net.

3COM, through its TippingPoint division is creating a vulnerabilities market. They offer compensation for vulnerabilities found, and seem to have a responsible notification process.

The security researcher offers his finding to 3COM, and they decide how much it’s worth. The researcher then choses whether to accept or reject the offer. Now, this isn’t a real market, as there’s only one legit buyer - I’d like to see this done in a broker scenario, offering the sale to the highest bidder (presumably including the OEM responsible for the vulnerability). This would put pressure on the vendors to get their software fixed before shipping.

3COM doesn’t state (at least in what I skimmed) if they charge the OEM for the information. Their financial motivation seems to be that their IDS product is more up-to-date than the others because they have first-crack at the information.

This process exposed a QuickTime vulnerability that’s exploitable through a web browser allowing Java to execute (presumably through the QuickTime for Java bridge). This is just an illustration of the fact that as you expose more of the outside of the sandbox, your risks increase. Shore up your sandbox with NoScript, which, incidentally, now offers some XSS protection. This was previously reported in the press as a Safari vulnerability, but it’s not - it’s exploitable through Firefox too.

The only downside to this whole process is that they’re not allowing blackhats to play. They want to be able to file the proper IRS forms, they say, but I can’t see why fees to blackhats necessarily cause a problem here, especially if the funds are wired off-shore. Extortionists are paid all the time by big online companies, and there are no IRS forms there. This is some kind of excuse on 3COM’s part, but I’m not sure what they think their exposure is by allowing the blackhats to play - their products would increase in value just the same, and vendors still get their products fixed. Right now, the blackhats have to turn to a Mafia to get money for their codez, so IRS or not, allowing blackhats in only improves the security situation, and perhaps turns a few blackhats into whitehats.

On Monday I went to a memorial service for Suzy Smith, a friend and former co-worker. Suzy and I chatted over e-mail regularly, mostly about new Mac products, but more recently about some treatments in clinical trials for curing metastatic lung cancer.

Suzy was a Mac fanatic, an appellation she’d happily accept. It was really quite a curious memorial service with so many stories about her love of the Mac, and it wasn’t just because the Chapel was half-full of IT department workers. There’s something interesting about it beyond Suzy - I can’t imagine a similar scenario where you could replace Mac with Windows or even Linux and get so many warm smiles from a group. Suzy would have had lots of reasons to explain why this was true. If I were faster thinking, I would have sent a bouquet of flowers stuffed inside a Compact Mac case (alas, I think of these things too late).

Tonight I had the sad duty of deleting Suzy from my Buddy list and her vCard from my Address Book. But it’s an OSX Address Book, so she’d be OK with it.

Her family asked callers to the reception to take a ring from Suzy’s vast collection. I picked one out for Emma with flowers, in three kinds of gold, and we’ll have that remembrence of her for many years to come. Her sister said they didn’t know what to do with a hundred rings. I reassured her that what they chose to do was great, making a hundred people happy, something Suzy would have liked maybe even more than a new iLife release.

Photo: two coworkers decided, independently, to make 6-colored Apple cookies for the reception.

I’ve recommended SuperDuper to a few folks looking at Mac Backup solutions based on this survey so I figured it was time I implemented it myself. I’m sorry to report results are poor.

For some reason SuperDuper gets confused when copying some files. I think there’s probably a race condition with transient files. It reports I/O errors on files that no longer exist when I look at them, and I/O errors on files that I can copy by hand, with cp, and with ditto just fine. It’s not open source, so it’s hard to say exactly why that happens.

Now, that, in and of itself, wouldn’t be a huge problem - I see the same kind of thing with rsync, and any other backup tool (but perhaps not with ZFS snapshots – but that’s another article). However, SuperDuper stops the backup process when it hits a few of these errors, leaving a potentially not-backed up disk, for a few file copy errors.

Since I was planning to purchase SuperDuper, I sent the developer a note about it. His built-in reporting tool is very smart about integrating with AppleMail and used the Keychain appropriately - a nice touch. He wrote back with a link to a blog article he wrote about this issue as a response, and assured me that not to worry, most of the files were backed up anyway.

His point is that if there’s an I/O error, then you have a disk problem and need to get that looked at. Now, he has a point - if there’s a real disk error his tool isn’t going to fix it. But, in this case, there’s no apparent disk error, so the behavior of SuperDuper is incorrect. Even if there were a disk error, I’d be trying to use a disk backup tool to recover the data. Sure, if you have $2K, send it off to DriveSavers - they’ll do a better job, but most people don’t.

But besides that, his assertion that most of the files were copied anyway was incorrect, by SuperDuper’s own reporting (and a visual inspection of the target disk image) – in my case, SuperDuper stopped after 553,643 of 1,987,938 files. So, not so complete.

Clearly, for my usage, SuperDuper wasn’t copying most of the files, and was stopping on false I/O errors. This makes for an unusable backup tool.

So, I wrote back to the developer, explaining what I had seen and asking if he could include an option to allow the user to decide whether continuing after I/O errors was the right thing to do (most tools - rsync, dd, etc. all have such an option).

This is the entirety of his response:

From: support@shirt-pocket.com
Subject: Re:(Case 38299) SuperDuper!: Error during copy - copy stops due to I/O errors
Date: April 10, 2007 18:51:38 EDT
To: bill@bfccomputing .com

I’m sorry this is a problem for you, Bill, but I wish you the best in finding a good alternative.

Dave Nanian
Shirt Pocket

Clearly, he feels so strongly that his behavior is the correct one that he’s willing to forgo many sales to not give the user a choice in the matter. I respect his conviction while strenuously disagreeing with his conclusion. So, if I’ve recommended SuperDuper to you previously, please consider that recommendation rescinded and accept my apologies for not testing it thoroughly enough first. I’ll try to e-mail a link here to everybody I can recall.

Now, that leaves the problem of what to use for Mac backup software. I have some Enterprise clients using a version of rsync I patched up. I’ve done some ditto backups locally. But all of those things have problems, and none are currently the right answer. rsync is probably the one with a chance, since it’s open source, but the patch sets to date are inadequate. If the entire problem with copyfile() is fixed in Leopard (perhaps via ZFS) then rsync can start working. I’ll be testing that once I’m confident I have a good backup of my laptop (a Catch-22, no?). Otherwise, it’s probably going to be time to start up a SourceForge project to get interested collaborators together to come up with a working solution.