[Opensim-dev] Supplying IScene instead of Scene for the future region modules mechanism

[Melanie]

The important turning point of the discussion was when the talk 
turned to splitting things out of Scene. I'm still dead set against 
hiding Scene as it is today behind an interface, because it would 
turn into just as unmaintainable a behemoth as IClientAPI is.
However, breaking things out of Scene into contained classes and 
making Scene into what it should be - a manager for the objects 
contained in it - makes that concept viable.

I often react emotionally and quite strongly, you know that of me. 
You also know I do come around, you've seen that many times.

Melanie


Justin Clark-Casey wrote:
> Melanie wrote:
>> Looks like you didn't read my post. I said, as a mid to long range 
>> goal, yes. I didn't say we should never have it. I said it would 
>> block critical fixes if it were forced onto the new region module 
>> interface now.
> 
> In your earlier posts you vehemently decried any notion of supplying a scene interface.  I hope you haven't forgotten 
> these already.
> 
>> I'm not bullying anyone. You have not yet had the pleasure of seeing 
>> me bully anyone. :)
> 
> And I've no interest in trying to discuss something with someone who uses extreme statements without any hint that they 
> can see the other person's point of view.
> 
>> 
>> Melanie
>> 
>> Justin Clark-Casey wrote:
>>> Sean Dague wrote:
>>>> Melanie wrote:
>>>>> Hi,
>>>>>
>>>>> as a mid to long range goal, +1, actually.
>>>>>
>>>>> but in the short run, the ability to load and unload regions is 
>>>>> blocked by the existing module API, and to fix this basic piece of 
>>>>> functionality, they need to be migrated to the new API, asap.
>>>>>
>>>>> If this is dragged into a long architectural discussion, we won't 
>>>>> get region restarts for many more months.
>>>>>
>>>>> So, I'd rather see this iteration of the region module API pass 
>>>>> Scene, and remove the old API very soon, and then think about 
>>>>> architecting and refactoring when that is not a blocker to 
>>>>> adding/repairing basic functionality.
>>>> I'd agree with Mel here about lets keep it a bit more open and sloppy
>>>> for now, and start to lock that down once we're on the other side of the
>>>> loader issue.
>>>>
>>>> We all come to this from different perspectives.  Mine is a lot of scars
>>>> and lost time due to IScriptHost a year ago.  Just about every LSL
>>>> commit required changing IScriptHost and adding back in functions for
>>>> SOP.  Eventually, I just threw out IScriptHost, as it was clear that
>>>> interface was far too premature.
>>>>
>>>> I think we're a bit premature on IScene at this point.  We know what we
>>>> all would do with it, but leaving the barn door open to other random
>>>> folks abusing the interfaces in ways that we didn't expect is probably
>>>> reasonable at this stage, so we make sure we don't lock off a piece of
>>>> function that's very reasonable to want.
>>>>
>>>> That being said, breaking Scene into more digestable parts would be a
>>>> *very good thing*.
>>> Definitely.  I've spent quite some time moving things out of Scene myself.  Much of what remains are the much more 
>>> indigestible bits (e.g. land/terrain management, inventory).  I plan to look at these in the course of my normal 
>>> attempts to chip away at the big ball of mud but any help here would be much appreciated.
>>>
>>> Switching the region module mechanism seemed to me to be a good opportunity to introduce a Scene interface without 
>>> causing a separate bout of pain later on.  But on hearing some of the rational feedback, I accept that it could still be 
>>> too early to do this.  If people were to treat this interface as a contract they could build against, as Teravus pointed 
>>> out, then it needs to be stable.
>>>
>>> Nonetheless, I'm seeing that with the exception of Melanie, the core developers who have posted are in favour of having 
>>> an interface eventually (I presume before 1.0).  So it is coming and if people have to endure some update pain later on, 
>>> well that's the price of building against alpha code.  It's not a huge upheaval either, in most cases it is simply a 
>>> search and replace of Scene with IScene.  When this happens later on I will point back at this discussion as warning.
>>>
>>> I'm quite happy to hear arguments against a scene interface, but I will forcefully counter and later ignore any attempt 
>>> by Melanie to bully me off the point with completely exaggerated statements.  We're here to discuss, negotiate and 
>>> compromise in good faith, not to shout at each other across the mailing lists.
>>>
>> _______________________________________________
>> Opensim-dev mailing list
>> Opensim-dev at lists.berlios.de
>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>> 
> 
>