[oi-dev] Moving on with postgresql

[Reginald Beardsley]

 Does the repository server collect information about installs? It seems to me it should and package and version support decisions based on that information. Without that information there is great potential to waste resources doing work which will never be used.

In the early 90's I was a big advocate for buying binary FOSS packages as compiling them for multiple platforms was not cheap unless it was done during idle time while waiting for another job to complete. I routinely did this on other systems while waiting for a compile and test to complete on the platforms we supported for our package.

As a practical matter, distribution of builds of FOSS was out of scope. I just did the 6 platforms which used our NFS server for our convenience and efficiency. No one asked me to do it. I just did it and everyone was very pleased that I did.

In order to better manage resource allocation of 3 people, I implemented a usage logger which collected usage data via UDP from all the business affiliates in a major oil company for the package we supported. Once a month I generated plots showing program usage by affiliate and cumulative usage worldwide. That allowed management to allocate resources to programs in widespread use and ignore programs only used by the scientist who wrote it. One individual was *very* noisy. It was extremely useful to be able to show he was the sole user and had run the program 1 or 2 times in the last year. As he had written it, he was quite capable of adding the "important" features himself.

I do not suggest collecting per use data, but per install data seems entirely sensible.

Reg
  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openindiana.org/pipermail/oi-dev/attachments/20210417/34d8b455/attachment.html>