[TNG] Status (and merging)

[Elrond]

On Mon, Jul 24, 2000 at 10:10:20PM +1000, Gerald Carter wrote:
> David Bannon wrote:
> > 
> > Does this mean that we are heading into the same dead 
> > end that stopped the 'old head branch' circa mid '99 ? 
> 
> Possibly.  We've all known this though, and I've even 
> said it on this list.  There's no reason to get back
> into the why this happened again.  If you've been 
> around long enough, you know why.  TNG was never meant 
> to become a release branch.  It was meant for Luke to 
> experiment with so that development in the main branch
> could continue and releases/bugfixes could continually
> be done.

How I see it currently:

TNG is a playground for Luke and others (including me).
Some of us try to merge stuff from HEAD. Luke did a great
job in this area... When the mods to HEAD aren't too
complex, I'm trying to merge them into TNG too.

(For example, the whole select/notify-change in HEAD was
too fast/big to easily follow it, so I just even didn't try
it.)


> > If TNG cannot be merged back into mainstream Samba 
> > then that sounds like mainstream Samba won't do PDC 
> > to W2000 and that is plain scary.
> 
> Nope.  That's not what i said.  We have to grab designs
> and code from TNG but certain things like 20 daemons
> to run Samba will mostly likely remain only in TNG.

Yeah, that's how I see it too.

BTW: Currently merging over the server-code for the daemons
and not using different daemons would create a big
security-hole.
I have some plans to fix this in TNG. But my time is
seriously limited.
(I've a whole lot of plans on internalchanges for TNG,
including this one and others to make live with sidlc lots
easier, most of these are nearly fully complete in my
head... I simply need the time to code them up...)


[...]
> Currently, the Samba team is made up of about 2 dozen 
> people.  So I'm sure that begs the question, "How does 
> one become a member?"  It's not hard really.  Write code,
> write documentation, take the ball an run with it (remember 
> that Andrew is the benevolent dictator in this).

:)

> Now out of two dozen members (grown in size over the years),
> code check ins now-a-days are by about 1/2 dozen people.
> That's a programming team of 6 people staring at bits on 
> the wire, reading as much documentation as is available
> (writing it in other cases), and still trying to maintainable
> code quality in releases.

and wasting their free time... (for... hmmm)


> Does anyone here realize that we had to rewrite the entire
> locking semantics for 2.2.0?  it is now what we believe to 
> be the most robust, and solid locking code available in 
> Samba or out. (ask me this again after the release of 
> 2.2.0 though :) ).

This locking-code is already in TNG, and people seem to use
it, so it can't be too broken. ;)

> Andrew has also implemented a small 
> database library and we've lookup stores to this in 
> order to improve speed and scalability.

Sidenote:

I've already written this up on samba-technical:
tdb doesn't scale properly: If the tdb gets too large, you
end up with linear searches, because the hash-tables have a
fixed size. (I know, how hard it is to get dynamic resizing
hashtables [in a file], so no offense intended!)


[...]
> So at this point I'm probably rambling.  I hope that no 
> one has taken offense as it was certainly not meant in 
> anyway.  I just think what has happened (and please don't 
> anyone take this wrong), is that to focus on the PDC
> implementation is but one part of the whole.  Maybe 
> everyone has got stars in their eyes from watching Luke 
> (and Elrond and Matty and others) work on this.

;)

I would like to see someone with stars in their eyes. ;)


[...]
> Until you actually look at the code in TNG
> and then look at the code in HEAD, you will **never** 
> understand what we talk about the differences between 
> the two.  Trust me on this one.  I know that which I 
> speak of since working on the rpcclient merge from TNG 
> into HEAD. :-)

I know some of this stuff... so I haven't looked much at
HEAD.

And I fully respect all the work on HEAD/2.x, because I
know, this work is done with attention to stability and
stuff like that, because that is, what all the major users
want (including me for non-pdc-stuff. I wouldn't tell
people to use samba, if it weren't that stable)
On the other side, I can work on TNG and don't need to
worry, when it breaks for some time, because everyone, who
uses it, knew before, that they're using alpha-code.

One of the reasons, why I've also currently stopped working
on TNG: If the current TNG-code looks to be stable to some
degree, Luke can make up an alpha-2.6-release and if
someone complains about cvs being broken again, one can
simply tell them to use 2.6. ;)


And to write up some "final words":

I don't want TNG to die too early. My idea is, that TNG
being something like "reference-material" for PDC-stuff and
HEAD slowly taking over stuff from it. When HEAD has
everythig from TNG, TNG might die.


[...]
> Please, no flames, ok?

I liked the mail very much, so why should I write a flame?
And I think, I tend to be more on the TNG-side. ;)

> I will gladly respond to logical questions comments, but
> not flames.  I'm way too busy right now.

:-)

> Oh...and thanks for using Samba. :-) :-)

Thanks for maintaining it. Without it, this small institute
here in the university either wouldn't have any pdc
(everyone with a local account) or a scarry nt-pdc...


> Cheers,
> jerry


    Elrond