August PERT Chart Reading

[Marc Stiegler <marcs>]

Well, the PERT chart is now so large and unwieldy that MacProject 
is almost unworkable. I shall have to break the PERT into pieces 
before I can make the next full-blown upgrade. Sorry it took 
so long to get this analysis out.

When the triage meeting ended, I was actually quite hopeful that 
we had gained back much of the time we had otherwise lost. But 
when the triaged parts were deleted, and the heretofore missing 
parts replaced them, they almost canceled each other out for 
Beta shipment.

Good news first, because the bad news requires so much discussion: 
the date for Alpha release of the backend has moved up a lot. 
In past PERT charts, the date of Alpha release was so close to 
Beta that it wasn't a useful milestone. Since we've sealed off 
orgls & berts, we should have Alpha this month. Alpha is good 
enough so that Greg Lutz and Raj Patel can get really serious 
about development, though they'll no doubt run into significant 
bugs and will still need patience and enthusiasm. (and as no 
minor side benefit, Alpha is good enough so that our own frontend 
developers, hugh and roger and ravi, won't be slowed down so 
much by an inadequate backend to hook into).

Now the bad news: Beta of the backend does not start until December 
27.

Of the three in-depth analyses of the PERT we have done so far, 
this is the second one in a row in which I have to report serious 
delays. Back in May I observed that we were staying on schedule 
only because we kept simplifying the design so much. The reduced 
size of the total project masked our inability to complete individual 
implementation tasks in the allotted times. I observed then that 
the challenge of keeping up with the PERT chart would rise relentlessly 
as the design became cast in concrete. I wondered whether we 
would be able to stay in place, running as fast as we could.

It has, and we haven't.

Reasons for the current slip:

1. It took us an additional 3 weeks, beyond the additional weeks 
we'd projected last time, to complete the document & links layer.

2. The slips in the development of URDI, which have been accumulating 
for a long time, have mounted to the point where the tasks stacked 
up behind it thrust out beyond the existing critical paths. I 
had not highlighted URDI slips in the past since they were far 
from the critical path. That is no longer true.

3. Reconsideration of the time allotted for some of these tasks, 
based on new wisdom from reviewing errors made in  our past 
estimates, caused us to make more accurate (usually longer) 
estimates.

4. A variety of new tasks, mostly small, have been recognized 
and added; the merger of the diverging branches of preAlpha development 
is one of the larger ones.

Hill is the critical path, by a nose. According to this PERT, it will take 
him about a week longer than either markm or dean. However, one 
bright spot in this report is that hill has been completing tasks 
faster than I can update the PERT, forcing me to add a new kind 
of correction box on the chart--the negative correction, where 
the task took a lot less time to complete than projected. So 
by the time I publish this memo, hill may have already knocked 
that extra week off his schedule.  

In practice, hill, markm, and dean are ALL on the critical path. 
Michael has ALSO gone critical, though his part of the chart 
has not been updated enough to show it (don't worry, michael, 
we'll fix that this week).

If hill does knock yet another week off his path, the official 
Beta date will move up by a week. But let us not get carried 
away by precision when accuracy is so much in doubt.
 
Let us look candidly at our predictions and results. We have 
made too many too-optimistic predictions to accept these dates 
at face value. We cannot sincerely assert that we will be ready 
on the day appointed by this PERT. 

Unless we can tap a hidden reservoir of strength deep within 
ourselves, we will not make Beta this year. I know there's no 
one here who does not feel a fierce frustration at this.

Folks, December 27 is 4 months, 2 weeks, away. 

>From time to time I have said we need the pacing of a marathon 
runner. We need to run as fast as we can but to still keep 
on running for a long time. 

Some marathon runners pick up speed as they come toward the end 
of the race. Now, we too need to find that kind of power, to 
excel beyond our best. It may be too early to try to transcend 
the determined yet steady pace. But it may be too late.

We need to accelerate past the ordinary marathon runner. We need 
to start sprinting.

I will be discussing with Chris Record additional measures we 
might take to avoid slipping into next year, or to at least minimize 
the duration of such a slip.

--marcs