[lwip-users] Re: [lwip] The future of lwIP

[John C. Toman]

Hi Kieran,

There are already a number of #ifdef options in lwIP, selectable through 
the lwipopts.h file, and I don't see a problem with adding a few more, 
as long as they're well documented and don't constitute large changes. 
For instance, I'm not sure if gigabit modifications are appropriate for 
this; perhaps there should be a separate gigabit code tree. But I don't 
see any problem with Paul's main premise of selecting features via 
#ifdef for lwIP's main targeted area (memory constrained 8, 16 and 32 
bit devices). Adam is already going down that path for lwIP 0.6 with 
being able to select the protocols to be included, for example.

The RFC 1948 TCP ISN stuff is an example of another potential #ifdef. 
But, then, as you say, it would only be a few lines of code and a couple 
of functions, not a radical change.

I also think that someone, or a very small group of people (lwIP 
committee?), needs to steer the lwIP ship, and the number of people with 
cvs write permission should be limited to a small number. The rest of us 
would then submit patches to the committee, who would decide whether or 
not to include them. This isn't really any different than now, except 
patches could be put into the code tree by someone other than Adam.

Regards,

John

Kieran Mansley wrote:

>On Mon, 19 Aug 2002, Paul Sheer wrote:
>
>  
>
>>>all of everyone changing lwIP to suit their own purposes.  For example,
>>>I've made many changes to my own branch of lwIP to make it more
>>>suitable for gigabit speed networking, but these changes would not
>>>be suitable for the main tree as it would be detrimental for many
>>>others.
>>>      
>>>
>>Myself I am interested in these Gigabit modifications, but
>>now I can't easily see them because they are not included!
>>    
>>
>
>If that's the case then maybe there should be a fork to allow people to
>diverge towards gigabit networking, but I wouldn't want to do that at the
>expense of people using small low memory systems, which is what lwIP is
>intended for.
>
>  
>
>>This is what #ifdef's are for. Most generic packages have many
>>many build options, and I would like to see LwIP become just
>>as flexible.
>>    
>>
>
>This is exactly what I'm afraid of.  #ifdefs are great for small changes,
>the odd line or two of code that needs to be different, but when you start
>using them to switch between radically different code structures it can
>get very messy and hard to maintain.
>
>lwIP is attractive for many people because if its simplicity and clean
>code, and the fact that it's easy to build and debug.  Trying to lever
>support for everyone's system into the one system would probably break
>that.  I'd much rather see it remain a generic stack, and have people add
>their own refinements.
>
>I think this illustrates exactly what I was trying to get at - we
>obviously have a difference of opinion about changes to lwIP.  In cases
>such as this where developers disagree, what happens?
>
>Kieran
>
>[This message was sent through the lwip discussion list.]
>  
>




[This message was sent through the lwip discussion list.]