The Perfect Thing, What is Xsan?, and Read Iris Pro
The Perfect Thing
I've been reading the book The Perfect Thing by Steven Levy, about the creation and growth of the iPod and Apple. My wife saw it at the library and grabbed it for me. Most of what Levy writes rings true to me, from my years at Apple. He quotes Steve Jobs about the coolness of Apple's products — and how Apple is not about cool. When Levy asked if Jobs he tried to make the iPod cool, Jobs responded, " No, we try to make it great." At a party one time another person said (in a very pejorative kind of way) to me something like, "Apple is very elitist." My response was, "No, it's not about being above others, it's about pushing our own limits." That's the essence of Apple — a company whose aim is to make great things. Not cool, not profitable, not good enough — it's all about great products. In a way, it's the reverse of the old saw, "perfect is the enemy of good enough." For Apple, good enough is the enemy of great.
Related to this was the part where Levy was describing some of the notorious Steve Jobs tantrums, where even A- work is considered not good enough, with the implied threat that unless you do A+ work next time, you're history. I would say that these stories are overblown. Why? Not because Steve Jobs doesn't do this sort of thing, but because the corporate culture is such that he rarely has to. Apple is full of people who reach the extra step, who routinely think in ways that are so far beyond the common that such tantrums are the exception, not the rule. I just hope I didn't drag down the average too much.
A last comment on this was the mention of the Apple 100 — the 100 most important people to Apple, according to Steve Jobs. They don't have the highest titles, but they're the ones you would save if the company was going under. This sort of organizational un-hierarchy is absolutely the way that Apple runs. Your job title and who you report to have little to do with your influence in the company. There are a number of people who are laws unto themselves, who can make things happen outside of formal channels. Where did I fit into that informal hierarchy when I worked at Apple? I'm sure as heck certain that I was never in any danger of being a member of the Apple 100. Was I in the Apple 1000? I doubt it. 2000? Maybe. 5000? Probably. (Apple has around 15,000 or maybe a few more employees these days.) There isn't a formal ranking system — but it's nice to speculate as to where I might have been on the scale.
What is Xsan?
After talking to a number of people at MacWorld, I've concluded that there's a lot of fuzziness out there about what Xsan really is, where it's useful, and what it isn't. I also sat in on a train the trainer class for Xsan last week, and it was interesting to see how the presentation emphasized the complexity of Xsan.
We need to start with an XServe RAID. This has two Ethernet ports that are used solely for managing the RAID -- none of the data on the disk gets moved through these two ports. It also has two Fibre Channel ports that are used to move data to and from the disks. When you want to attach an XServe RAID to a Mac (either a tower or an XServe), you need to install a Fibre Channel host adaptor card. You now have two choices.
First, you can attach the RAID directly to the XServe. This is called direct attached storage -- the RAID is functioning in a way that is little different from an external FireWire disk.
Second, you can attach several RAIDs to a Fibre Channel switch, and then attach several Macs to the Fibre Channel switch in an arrangement called a Fibre Channel fabric. There's one little fly in the ointment here. Each Mac will then try to mount all of the RAID volumes. This will lead to a clash, race conditions, and data corruption. You can alleviate the situation by using zoning at the switch, in essence making certain volumes visible only to certain Macs in the Fibre Channel fabric. One RAID volume is still connected to one Mac, but it's a little bit more flexible since you can unmount a volume, change the zoning, and mount it on another Mac. The built-in IP failover using heartbeatd and failoverd works by using this method to re-allocate storage.
Xsan takes the next step by allowing a RAID volume to be shared across multiple Macs on the Fibre Channel fabric. It creates volumes that can be mounted by *all* of the Macs on the Fibre Channel fabric. They all see the same files and data and all can access any of the RAIDs that are a part of the SAN. There are a lot of details in terms of tuning the SAN for maximum performance and making sure that everything stays up, but that's Xsan in a nutshell.
Read Iris Pro
Speakers got a copy of Read Iris Pro as a part of the speaker bag from MacWorld. I tried it out for a few hours and here are my thoughts:
Installation
I tried this out on an iMac G5/2.1 GHz, 1 GB RAM, MOSX 10.4.8 with Read Iris Pro 11.5.6. Installation was straightforward using the standard Apple installer. I give them points for not using lame Vise or Stuffit installers. All of the application was installed in /Applications/Readiris Pro 11.5/ except for a couple of fonts that were installed in /Library/Fonts. It is a universal application, so you will get native speed on Intel hardware. On the first launch the application presented a slightly confusing dialog presented about associating file types with apps (e.g. RTF with TextEdit).
Basic Test
I worked as a local admin user to start off. The Read Iris Pro software picked up the TWAIN drivers for the Epson CX-11NF multifunction device right away. A simple scan and recognize (one page, all text, some boldface, black & white) worked OK, the recognition went through and the RTF document was send to TextEdit. The OCR quality overall was quite good. Some examples of text that was mis-read are:
"10.4.X" was mis-recognized as "10A.X"
"afp://woserver3" was mis-recognized as "afp :/lwoserver"
"OD Master" was mis-recognized as "00 Master"
Complex Scan
I tried a more complex scan using four pages that contained text, handwritten notes, graphics, and colored text. To make it harder, I deliberately placed the pages in a document feeder so that they fed in at a slight angle. The first time I ran the test the application crashed. On second and subsequent runs the application was stable, so it appears that this was a fluke. The OCR quality was quite good as well. The document format (including boldface, font sizes, and text placement) was in general well-preserved. The handwritten notes did not OCR well at all -- they came through are random letters. In this run, the most common error was that lower case letters were recognized as upper case versions:
"appOXimately"
"eXisting"
Test From a Network Account
As a test of how well the application is engineered, I tried the four-page complex scan from a network user account with a network home directory. The application worked the same as it did for the local administrator account. I had to re-enter the serial number and was presented with the same slightly confusing file type dialog, but other than that everything worked as expected.
Conclusion
ReadIris PRO is a good OCR program for straight text, but technical terms with odd punctuation and formats tend to confuse it. The OCR speed was good, and it handles multiple page documents without problems.
--Paul