Date: Dec 6, 2012 4:56 AM
Author: John Fultz
Subject: Re: Pet peeve about version numbers of Mathematica
Actually, in this case, a revision number would *not* have been free.
If we had incremented the main version number of the product, it would
have meant recompiling from source many of the binaries in the product
to reflect that change. Which would have in turn required significantly
lengthening the test cycle. Which means that we could have been waiting
until close to the end of the week before delivering a patched installer
in the best case scenario, even though the only change we really wished
to make was to the installer.
The installer build went through our automated build system, produced a
complete log of the build, and was assigned a unique build number. That
build number is available in SystemInformation, since that is recorded
as a simple text file which is rolled up into the installer at the time
the installer is built. Additionally, we provide an MD5 hash, which was
updated when we posted the new binary.
I'm guessing your engineers haven't actually looked at Mathematica. I
just checked, and the Windows version has nearly 20,000 file system
entries (files and directories) in the installation. It boggles the
mind or my mind, anyway...to wonder how any team of humans could keep
track of 20,000 assets with tens of thousands of more source code files
without proper version control and build systems.
User Interface Group
Wolfram Research, Inc.
On Dec 5, 2012, at 2:17 AM, "McHale, Paul" <Paul.McHale@excelitas.com> wrote:
> I recently installed Mathematica 9.0.0 and had the problems. This was quit
> e unpleasant but I work with software people who understood. Things happen
> . They asked for specifics, I explained I upgraded from 8.0.4 to 9.0.0 and
> the problems were severe. Uninstalled 9.0.0 and the problems were gone.
> Explained about the problem as WRI explained it. All is good.
> A few days later the "new release" is out. Here is a conversation:
> Me>> So, the new release of Mathematica is out. Should fix the problem. Will probably upgrade next week.
> Software lead>> What is the new release number?
> Me>> 9.0.0
> Software Lead>> Wait (checks notes), OK, you mistakenly reported that as the problematic version.
> Me>> That is correct.
> Software Lead>> So you are changing windows to fix the problem?
> Me>> They updated the release package but did not increase the revision number.
> Software Lead>> If they are releasing two different items with the same reision number, what is their revision control? More of a guideline? (Pirate
> of the Carribean)
> Industry standard is you have to release different things with different revision numbers. As far as my boss is concerned, WRI appears to be non-programmers writing software. Not because of testing or bugs. But because revision numbers are free. Failure to follow propper version control sets off
> alerts that there may be greater issues below the surface. Remember, my Mathematica copy competes with his C# programmers. Revision control is not
> a trivial thing in defense engineering. Instead, it is yet another metric.
> Paul McHale | Electrical Engineer, Energetics Systems | Excelit
> as Technologies Corp.
> Phone: +1 937.865.3004 | Fax: +1 937.865.5170 |
> Mobile: +1 937.371.2828
> 1100 Vanguard Blvd, Miamisburg, Ohio 45342-0312 USA