Leak detectors
Overall, we are pleased with your evaluation of our runtime error-detection system, Sentinel I, in your review comparison, "Leak detector shoot-out," October 1994. We have since announced and shipped to beta customers Sentinel II, which catches the errors that
Sentinel 1.5 did not, namely uninitialized memory read and dangling pointers. Moreover, Sentinel II catches bounds over- and underwrite for global data, a capability that appears to be unique.
We would have preferred that you concentrate more on the usability aspects of the product. Three aspects we feel are extremely important are availability, licensing, and link-time processing.
Sentinel, as you mentioned in the article, is available for almost all common Unix development platforms. We cannot stress highly enough that if you are writing an application for multiple platforms, you need to test that application on all those platforms. Sentinel offers a huge advantage in this area that no one else can match.
Sentinel has extremely flexible licensing. An important note is that Sentinel licenses only the link step. So, Sentinel infrequently requires a license, and Sentinel users are free to move their Sentinelized software to any machine for testing, debugging, and beta deployment. This is a major purchasing discriminator for Sentinel.
Finally, we are dismayed that you mentioned Purify's support of threads in 3.0 but failed to mention that we support lwp on SunOS and DCE threads on Solaris 2 and HP-UX. We certainly support threads and stand ready to support others as demand warrants.
Ed Matthews
AIB Software
Dulles, VA
edm@aib.com
Most of your readers are probably either using or planning to use C++, which is why Purify has supported it for two and half years and established an estimated C++ user base of more than 10,000. Supporting C++ is a critical feature for products in this category, and this wasn't even mentioned in your review.
I also know users really care about performance and most commonly think of it in terms of build- and runtimes. Because of this, Pure Software has worked hard to make Purify as absolutely fast as possible. Your review did not reflect this. You
measured performance in
a completely different way than day-to-day developers see it. In fact, if you ran both Purify and Insight on Solaris 2.4 or on SunOS you would have found the performance numbers for your test case, xv (many times faster for Purity than for Insight).
Reed Hastings
Pure Software
Sunnyvale, CA
hastings@pure.com
Better and better
Your magazine used to take me 60 seconds to scan and toss. The December issue has cost me 45 minutes
already, not to mention the articles I've cut out to read later. Congratulations, it's getting really interesting.
Art Altman
Electric Power Research Institute
Palo Alto, CA
aaltman@epri.com
Debunking adb
Chuck Musciano, in his column "Debunking debuggers" (October 1994), has it right: ObjectCenter is a powerful development environment... However, he claims that it's too much trouble to use "complex" techniques like compiling with -g or using
a debugger other than adb.
While adb is Chuck
Musciano's perfect debugging platform, we don't think this is the case for most software engineers. Those of your readers who are not
developing applications in assembler should evaluate CodeCenter or ObjectCenter. We think they'll find it well worth their attention.
Chuck Musciano replies:
I agree with everything
Dr. Stacy notes about CodeCenter and ObjectCenter. The point of my article
was that adb offers certain
features (speed and ease
of access) that are currently missing in some of CenterLine's tools. I do use CenterLine products to track those tough pointer bugs, but lament the high cost of entry to get my code into the tool, ready to debug. I'm not alone in this; many developers I know lament the long start-up times of CenterLine products. My column was
a plea for tools that offer
the feature sets of tools like CodeCenter and ObjectCenter with the quick start-up and ease of accessibility that tools like adb afford.
Dr. Webb Stacy
Director of Engineering
Programming Environments
CenterLine Software
Cambridge, MA
Career denial
I keep up with many
computer periodicals, and want to let you know how valuable and personal (as
in "speaking directly to me") Edgar Saadi's Help Wanted (Career Advisor) column for August 1994 was.
"A"
Houston, TX
Correction
We welcome letters, preferably by
e-mail, to editors@advanced.com. Letters may be published unless sender notes otherwise, and may be edited for length and clarity. If you wish to subscribe please call 800-685-3435, or send address changes to Advanced Systems, P.O. Box 41534, Nashville, TN 37204-1534.
In our December C++ comparative review, we neglected to credit Plum Hall Inc. for providing the Advanced Systems Test Center with Plum Hall's C++ verification suite. The company is located in Kamuela, HI and can be reached at 808-882-1255 and 808-882-1556 fax. The editors regret this omission.
If you have problems with this magazine, contact webmaster@sunworld.com
URL:
http://www.sunworld.com/asm-02-1995/asm-02-letters.html.
Last updated: 1 February 1995.