[Ns-developers] GSL for ns-3
tomh at tomh.org
Thu May 28 11:46:39 PDT 2009
>From: Mathieu Lacage [mailto:mathieu.lacage at sophia.inria.fr]
>Sent: Thursday, May 28, 2009 10:26 AM
>To: 'Tom Henderson'
>Subject: Re: [Ns-developers] GSL for ns-3
>On Thu, 2009-05-28 at 18:12 +0000, Tom Henderson wrote:
>> Any support for or against including GSL as a required ns-3
>It really all depends on the amount of pain involved in making it
>possible to still use ns-3 if gsl is not available and, in this case,
>gracefully degrade the user-experience by disabling features. For gary's
>patch, the pain looks small enough that it's worth making gsl support
>I don't know about other features but, I really think that it's
>important to make the default build just work with as few dependencies
>as possible: there is nothing more painful for a newcomer to figure out
>he needs to install new packages just to get started so, I feel that
>it's important to make the default "./waf" command run smoothly as much
>as possible. We went through quite a bit of pain to make this possible
>so far, so, it would seem like a serious step backward to give up on
One additional argument in support of what you say above is that some users may want to run a large number of simulations on a university cluster (e.g. in advance of the paper deadline) and may not have ability to package upgrade those machines.
So, maybe a brief "HOWTO add GSL to your ns-3 model" would be sufficient at least for now, and ask each GSL user who submits to the main tree to figure out a policy for handling the case where GSL is not available. I guess you are fine with how the preprocessor is used to conditionally compile these models (since you pushed the patch), and otherwise it could be controlled by waf-- we can put examples of both.
More information about the Ns-developers