[Os-project-managers] [Fwd: OSoL, OSrL, change sets 2819 and 3013]

[Kipp Martin]

Hi Guys:

Before addressing the issues below here is some history.  This issue 
goes back to an early Jun-Kipp battle before Gus came on board. After 
doing OSiL we moved on to OSrL. I wanted to keep the numberOf stuff 
required since I was working with arrays in C++ and needed to know the 
number of things I was reading before reading them so I could do a new. 
Jun did not. Jun won. I was forced to go these as optional and then 
because I was working with C++ forced to use Vectors and not arrays. 
Later when this issue was revisited we changed. Yea!

Is the following accurate?

OSiL -- we currently have numberOfXXX required
OSrL -- we currently have numberOfXXX required
OSoL -- we currently have numberOfXXX optional

So Gus, if I understand your proposal, it is basically to go back and 
have 2/3 optional instead of 2/3 required. Is this correct?

I prefer we instead make OSoL required. Here is my logic. If we use 
optional instead of required we impose a tax on people who want to use a 
language such as C for implementation. In C there is no vector 
construct. So if you don't know numberOfXXX this is a big problem. What 
is the logic of imposing this tax on developers?

Gus, my vote is just the opposite of yours, my vote is to make the 
numberOfXXX in OSoL required just like in OSiL and OSrL. This has always 
been a strong preference on my part for the reason above.  I do NOT 
remember agreeing to make them optional in OSoL. I must have been out voted.

Cheers


> 
> The point of departure is the OSrL parser, and I (re?)discovered an
> important discrepancy between OSrL and OSoL. In OSrL we require the
> numberOfVariables, numberOfObjectives and numberOfConstraints
> attributes (but we allow them to be set to -1), while in OSoL they are
> optional.
> 
> So I did some digging. In revision 2819 we made them required in /both
> OSoL and OSrL/, and in revision 3013, we dropped back to making them
> optional in OSoL --- but not in OSrL. This is clearly inconsistent,
> and I can't exactly reconstruct the reasons for a) making the change
> in 2819 b) reversing it in 3013, and c) not doing it in OSrL as well.
> 
> Any ideas?
> 
> For the record, my vote would be to make the attributes optional in
> OSrL also, since we require numberOf... in every IntVector that might
> have anything to do with the solution, and since "-1" strikes me as a
> little odd as a value for numberOf...
> 
> Cheers
> 
> gus
> 
> 


-- 
Kipp Martin
Professor of Operations Research
and Computing Technology
Booth School of Business
University of Chicago
5807 South Woodlawn Avenue
Chicago, IL 60637
773-702-7456
kmartin at chicagobooth.edu
http://www.chicagobooth.edu/faculty/bio.aspx?person_id=12825325568
http://projects.coin-or.org/OS