A fair software distribution model?

[rskr]

I have been writing a software analysis package, in my spare time, and I want to distribute the package, after I have tested it on a number of different platforms and with a number of data sets, which I have invested in some compute resources to do.

Since, I haven't had the luxury of grants for support I am thinking that at some point and time I will want to monetize the the package, however for the time being I think I want to focus on producing a great product, and minimize overhead because I am competing with companies like Partek and CLC, and part of my business model is that I can produce and distribute software with less overhead. Though I do want to avoid the trap of having to compete with my own previously released products, so I am thinking that the initial releases will have to have a kill switch, since for clock based expiration I have seen bio-angels do things like reset bios clocks to run Genespring after their license expired, it will have to be a ET phone home model, however I will probably make it so that it just won't do new analysis, and previous analysis will continue to open. Does this seem like a fair way of doing business?

[ffinkernagel]

You just excluded anybody being serious about reproducability from using your software.
Not only is it a black box, no now it's a black box guranteed to stop working at some uncertain point in time.

Forensic bioinformatics is not much fun. And it's even less fun when it's your own project from 3 years ago that you can no longer reproduce.

And with reproduce I mean bug for bug, bit for bit the same output from the same input.

[rskr]

You just excluded anybody being serious about reproducability from using your software.
Not only is it a black box, no now it's a black box guranteed to stop working at some uncertain point in time.

Forensic bioinformatics is not much fun. And it's even less fun when it's your own project from 3 years ago that you can no longer reproduce.

And with reproduce I mean bug for bug, bit for bit the same output from the same input.

If someone did an analysis with my software, which isn't stochastic, what biologically relevant reason would the software not reproduce the exact results bit for bit, as the ones they would be able to continue to view anyway?

[ffinkernagel]

Your software won't run anymore, because you pulled the plug.

Biologists will routinely revisit data that was generated a few years ago and now want to either fiddle with parameters or analyze newer data in exact the same way (for example because the paper finally got into review and reviewer 3 requested such).
First step in such a setting is to verify that you can still get from input to the same output.



Personally, I don't endorse anything without source code for scientific research - because I need to be able - at least have the possibility - to understand the the methods used en detail, and I have yet to be able to re-implement any method with bit-for-bit output just from the paper.
I do acknowledge your desire to make money of your work - if I had to strike out like that, I would consider distributing the software as open source and charge for consulting / support / changes. That's a fair deal for everyone involved, you get to eat, I get my reproducibility, the software does got a decent user base according to it's merit, the users get what they need, and are not foobared if you were to drop out of developing the software.

[rskr]

Your software won't run anymore, because you pulled the plug.

Biologists will routinely revisit data that was generated a few years ago and now want to either fiddle with parameters or analyze newer data in exact the same way (for example because the paper finally got into review and reviewer 3 requested such).
First step in such a setting is to verify that you can still get from input to the same output.



Personally, I don't endorse anything without source code for scientific research - because I need to be able - at least have the possibility - to understand the the methods used en detail, and I have yet to be able to re-implement any method with bit-for-bit output just from the paper.
I do acknowledge your desire to make money of your work - if I had to strike out like that, I would consider distributing the software as open source and charge for consulting / support / changes. That's a fair deal for everyone involved, you get to eat, I get my reproducibility, the software does got a decent user base according to it's merit, the users get what they need, and are not foobared if you were to drop out of developing the software.

The problem with open source is reproducibility. Quite frankly most bioinformaticians couldn't even compile my code(They only know java and perl), and I have no way to control the libraries that my code depends on, I suppose that is where your consulting fees would come in. My approach to the special cases of extra support IE, someone needs the software to do something it wasn't specifically designed to do for example if it were unable to connect from behind a corporate firewall, because the port was blocked, was to provide some sort of support license, this could include doing new analysis using out of date software when appropriate.

[ffinkernagel]

Quite frankly most bioinformaticians couldn't even compile my code(They only know java and perl),

Do not underestimate your customer base.

[rskr]

Do not underestimate your customer base.

Who is underestimating anyone , bioinformaticians can't or won't program in c++, its too hard, they don't learn it in school, and they don't see any reason to do it when they are just going spend most of their time interpreting results and running programs anyway, not to mention management is more interested in being able to find cheap replacements and whether or not they "like" them.

[rhinoceros]

Who is underestimating anyone , bioinformaticians can't or won't program in c++, its too hard, they don't learn it in school, and they don't see any reason to do it when they are just going spend most of their time interpreting results and running programs anyway, not to mention management is more interested in being able to find cheap replacements and whether or not they "like" them.

I'm one of those that didn't bother to learn c++ but I'm curious why you say that it's harder than bash, python, perl of java. Looking at the basic syntax, I don't really see much difference between any of these languages.

I agree with pretty much everything ffinkernagel wrote. As to a fair software distribution model, I think Robert Edgar does it right.

[rskr]

I'm one of those that didn't bother to learn c++ but I'm curious why you say that it's harder than bash, python, perl of java. Looking at the basic syntax, I don't really see much difference between any of these languages.

I agree with pretty much everything ffinkernagel wrote. As to a fair software distribution model, I think Robert Edgar does it right.

I don't believe C++ is harder, in the sense that it is easier to do more challenging things with C++ if this makes any sense, though many people(the bioinformaticians who think C++ is hard, to whom I was referring) don't appreciate the mixture of rigid syntax checking and numerous runtime traps, which allows power users to specify how things get compiled very specifically and create efficient code, as well as manage consistency in larger code bases.

Robert's CV is interesting, I would wager that he didn't get rich at Parity Software with open source, though I can't seem to find any information on that.

[FrMartin]

I don't believe C++ is harder, in the sense that it is easier to do more challenging things with C++...

Correctly. It shouldn't be that hard if you're going to do some more advanced things.