Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!decvax!decwrl!amdcad!amd!intelca!oliveb!hplabs!hplabsc!taylor
From: taylor@hplabsc.UUCP
Newsgroups: mod.comp-soc
Subject: Re: Performance Monitoring
Message-ID: <723@hplabsc.UUCP>
Date: Tue, 7-Oct-86 12:59:22 EDT
Article-I.D.: hplabsc.723
Posted: Tue Oct  7 12:59:22 1986
Date-Received: Wed, 8-Oct-86 04:25:01 EDT
Reply-To: hplabs!harvax!cdx39!jc@hplabs.HP.COM
Organization: Hewlett-Packard Laboratories
Lines: 36
Approved: taylor@hplabs
Reference: <681@hplabsc.UUCP>


This article is from harvax!cdx39!jc
 and was received on  Tue Oct  7 00:51:31 1986
 
There's at least one situation where I wish some good monitoring-type 
statistics could be collected.  This is the general subject of programmer 
productivity.  We have gone through a decade or two of all sorts of 
unsupported claims that approach X (i.e., Structured, Modular, Top-Down, etc.) 
produces perfect, user-friendly, maintainable, etc. software.  Rarely is any 
attempt made to scientifically test the claims. 

Regardless of the religious aspects of the discussion, I'd really like to see 
some good data.  Is strong typing helpful?  It seems intuitively that it 
oughta be, and C has been evolving in that direction.  But I suspect that in 
fact it isn't.  (Counterexample:  I've used Snobol4 enough to be fluent in it; 
it has no type declarations at all; it is one of the easiest languages I've 
ever used.)

One thing I've noticed in my own experience is that when I have 3 or 4 
terminals available simultaneously, I can debug code much faster than with 
just one.  Or at least I think I can; I don't quite have any hard data to back
it up, and by myself I can't very well conduct a blind test...

If my experience is correct, then it would be a good investment for any 
software-development company to buy several terminals (and associated lines 
and memory) for each programmer.  The cost of the hardware is small, and would 
be paid back in a week or two.  I've found that it is impossible to convince 
people of this by argument; what we need is good, solid data that can only be 
collected by some sophisticated monitoring.  I'd be overjoyed to partake in 
such studies, if they would lead to improving the work environment.  (I don't 
like having to debug comm software with only a single terminal on my desk; I'd 
like to be able to hand my boss some published studies and demand 3 more 
terminals.)

[moderator comment: I'm not sure we really want to go down the road of
 what is good programming, and so on, but the question of HOW to monitor, in
 an unobtrusive way, is something I'll discuss further in a bit...  -- Dave]