Re: Moving forward

[Aaron Turner <aturner@linuxkb.org>]

On Mon, 5 Jun 2000, Robert Myers wrote:

> On Sun, 4 Jun 2000, Aaron Turner wrote:
> 
> > I'm a bit confused... didn't Jason tell you?
> 
> 	Yes, he did.
> 
> > The reason for trashing the current site and code boils down to one basic
> > problem: scaleability
> 
> 	Isn't this a problem that could be solved by re-coding the site at
> a later date?

Well that was the only way to solve it.  Problem was that nobody was
really interested in writing code that we knew:

1) We'd throw out soon
2) Wouldn't solve all the problems (performance issues)

> > The code was not scaling in terms of modularity and performance.
> > functions.php3 was almost 1,500 lines of code, was poorly commented, and
> > difficult to manage.  Furthermore, a key function which is a part of
> > nearly every page had a 7-8sec runtime on a PIII-733Mhz.  Our analysis of
> > the problem indicated that it was a problem with PHP itself- optimizing
> > the code would have little effect.  Moving to PHP4 did not look like it
> > would resolve the performance problem completely and it would have no
> > effect on the modularity issue.
> > 
> 
> 	I talked with Jay about this a little bit, he said that this was a
> result of the comments section for the articles.  Wouldn't it have been
> prudent to move that functionality to a later time frame of the project?

I don't rememeber Jason mentioning that we could easily strip out the
offending code.  My understanding was that it comprised a major portion of
functions.php3 and pages that didn't have comments were also effected.

> > And the modularity of the code was causing Jason tremendous problems with
> > fixing bugs.  It also made it nearly impossible for anyone other than
> > Jason to actually work on the code.  Frankly, nobody but Jason really
> > understands the code though I know you and Daniel understand parts of it.
> > Not to mention, nearly impossible to add any more developers to the
> > project in the future.
> > 
> 
> 	Well.  I agree.  Documentation was the last thought on anyones
> mine.
> 
> > Which brings us to the decision to bag it all.  Basically, we (Jason
> > really) could fix the bugs and add some basic functionality in about 2
> > months.  It would not however fix the 7-8sec execution per page; which
> > would render the site basically unuseable, nor did we have a real solution
> > for this without *major* code re-write which would take significantly
> > longer.
> > 
> 
> 	See my note above about the comments portion.  7-8 seconds is
> definately unacceptable.
> 
> > Jason has for a long time wanted to re-write the backend using CORBA, and
> > we decided that we could either start on this now or wait a few more
> > months.  After looking at our options, it appeared that waiting any
> > further would be wasting our time on "dead code".  I want to make it clear
> > that Jason, Daniel, and I were all in agreement in this and everyone had
> > took a few days to think it over before we decided on a course of action.
> > 
> 
> 	The code is not really dead, and the time is not really
> waisted.  The program logic could easily be moved to another
> language.  This is really all a part of a program lifecycle, as new
> technologies, and new ideas emerge, the old code is really either thrown
> away or whatever.

True.  I was refering to it as "dead code" because everyone felt that it
was a dead end.  The real issue was did we want a working site while we
worked on the rewrite.

Everyone wanted a functional site during the rewrite, but at the same
time, nobody felt it was worth 2 months of effort- when we could spend
those 2 months on the new backend instead.  Unfortunately, we just don't
have the manpower to do both at the same time.

> > It also has another major benifit- OODS is applicable to a lot more
> > projects than just the LKB.  This means we should find it significanly
> > easier to find developers interested in contributing.  CORBA also allows
> > us to utilize a plethora of languages which should prevent simular issues
> > to thoses of the PHP performance issue.
> 
> 	Yes, I know about OODS.  And I also know Corba will help out the
> project more.  I am still confused about why it is just all dumped like
> that.  It seemed as though the last time I worked with Jay on the code,
> that you guys were so close to having something, albeit not the grandious
> (sp) thing that Jay wanted it to be, but something none the less.

We were gravely mistaken.


--
Aaron Turner, Core Developer       http://vodka.linuxkb.org/~aturner/
Linux Knowledge Base Organization  http://www.linuxkb.org/
Because world domination requires quality open documentation.
aka: aturner@vicinity.com, aturner@pobox.com, ion_beam_head@ashtech.net
The difference between `Unstable' and `Usable' is only two characters: NT