Bug Watch: Don't Feed the Bugs

Bother with boundaries, distinctly dastardly dialogs and a couple of contrasting comments.

Fix: Autodesk has distributed hotfixes to address this problem. Some people report that these hotfixes correct the above errors, but other crashes still occur. Keep a close look on the Autodesk Web site for breaking information about this problem.

Boundary Bother (R12 to 2007)
One thing that has annoyed AutoCAD users for years is the boundary detection mechanism in AutoCAD. This mechanism can be used on its own, but it's usually invoked during hatching. AutoCAD frequently fails to correctly detect what would appear to be a fairly obvious hatch boundary. Sometimes zooming in or out helps, sometimes it doesn't. People frequently must resort to drawing on top of existing geometry to get their hatches done. I know some users who use MicroStation as a workaround when their drawing refuses to hatch in AutoCAD. It seems capable of easily handling what AutoCAD finds so difficult. Autodesk has tinkered at the edges of this problem for years, but it's about time it was fixed properly. You can help.

What can you do? Every time you come across an area that AutoCAD refuses to hatch that you think it should be able to, report it to Autodesk. Wblock out the offending portion of the drawing, confirm that the problem still occurs in the Wblocked drawing and submit a bug report. I don't know if you have recently tried to look for the bug report form on the Autodesk Web site, but it's not obvious -- even if you do a search. To save you the trouble, click here.

A Lack of Dialog Revisited Again (2005 to 2007)
As I reported in February and March, sometimes AutoCAD's dialog boxes just don't want to play by the rules. Judging by the number of e-mail messages I get about this subject, it is one of the most (or should it be least?) popular bugs out there. I have had reports of the first sightings in AutoCAD 2007, so it looks like Autodesk didn't manage to squash this one. It will be quite difficult for them to do so, as it doesn't usually crash AutoCAD and therefore doesn't generate an error report. It also doesn't want to make an appearance in front of Autodesk people. I've never even seen it myself. What a shy little critter!

Some people report that they can detect it creeping up on them. When the Xref notification bubble starts losing bits of text, the bug is on its way. When the bubble contains nothing but an X, it has arrived.

Comment: No Version Aversion Here
In June 2003, I berated Autodesk for AutoCAD 2004's removal of the ability to save in both Release 13 and 14 formats. I felt that removal was a bit too early for the market. Now, it's time for some praise. AutoCAD 2007 restored the ability to save in Release 14 format. Although that ability may be rarely required these days, it is good to see Autodesk expanding, rather than contracting, the backwards compatibility of its programs. AutoCAD 2007 is the most compatible release in AutoCAD history, capable of saving in the four native formats of eight AutoCAD releases, covering a period of nine years. Well done, Autodesk.

Comment: Is a Year Enough?
I will be examining AutoCAD 2007 for insects over the coming months, but first I will examine a more philosophical question. Is a 12-month product cycle good for AutoCAD?

In the distant past, AutoCAD did not have regular release intervals. Instead, the software was developed until it was considered ready and then it was released. Now, Autodesk has successfully disciplined itself to producing a release at almost exact one-year intervals. This schedule is great for Autodesk, because it means an effective doubling of upgrade revenue when compared with the days when it was roughly two years between releases.

The yearly upgrade was attempted from AutoCAD 2000 onwards. The first couple of attempts, 2000i and 2002, suffered from a lack of compelling features compared with what people had come to expect in a full-price upgrade. I remember one wag commenting that the "i" in 2000i stood for "ignore." Autodesk learned that it must continue to cram in enough useful content to attract people's business.

AutoCAD 2004 had a longer gestation period and contained more useful stuff. Of course, all those people on subscription then complained that they had paid for a year's worth of upgrades in which they received nothing. As Autodesk was (and is) encouraging people sign up for a subscription, that didn't work so well. Since then, Autodesk has stuck to its guns and has shipped a new version every year.

I've been impressed by how many useful new features Autodesk has managed to add during this shortened timeframe. I've been less impressed by how well finished those features are. A prime example is the CUI, but there are many others. The CUI was clearly not ready for release, but Autodesk inflicted it on us anyway. Not good. Even less impressive has been the lack of CUI progress made during the subsequent year. Some things are better, but it's still not fully cooked. Look out for more on that in coming months.

One of the biggest problems with releasing a partially baked feature is that it tends to remain partially baked for years to come. Autodesk must not only fix the problems, it must also do so in a way that avoids inflicting further problems on existing users. All of which makes the job considerably more difficult than it would be if it hadn't been released yet.

Autodesk releasing unfinished work has often been a problem, but the 12-month cycle encourages it. In retrospect, 12 months wasn't enough time to fully develop a well-designed and well-implemented new menu system, but Autodesk tried anyway. By the time it was clear that CUI wasn't going to be ready on time, it was too late to remove it from the product. If the release date is not considered moveable, then quality is going to suffer.

Another issue is that Autodesk must try to find time to fix last year's problems at the same time that it's trying to write and debug this year's new features. Because of the short and fixed release schedule, there is no room to move. Either the new features suffer a lack of development, or the old ones continue to suffer.

Of course, it does not follow that a long gestation period necessarily leads to a reliable product. The notorious Release 13 had the longest gestation period in history. However, Release 13's problem was not that it was released too late. On the contrary, it was released too early. The shareholders were getting nervous, and Autodesk decided to ship the product, ready or not.

Almost the same thing is happening today. Then, the shareholders were temporarily placated by shipping unfinished software as a panic reaction. They now are being permanently preemptively placated by shipping unfinished software according to a fixed schedule.

In conclusion, I would have to say that the fixed yearly release schedule is not good for AutoCAD. It is good for Autodesk, certainly in the short term, but that's not at all the same thing as being good for AutoCAD or its users.

What can be done? Autodesk is not going to want to move to an 18- or 24-month cycle and still charge the same amount for an upgrade. If you changed your business practices to successfully double your revenue stream, would you change things back again? That's not likely to happen. If you want AutoCAD quality to improve by extending the release cycle, you have to be prepared to pay for it. If you're on subscription, you would have to be prepared to go without an upgrade during a subscription year. If you purchase upgrades, you would need to pay 50% or 100% more for each release, so you pay the same total amount of money for fewer releases over a period of several years. Are you prepared to do that?

Let's hear what you have to say on the subject. How long do you think there should be between AutoCAD releases? What do you think about paying more for a release? Send your viewpoints to me at editors@cadalyst.com.