Subject: Re: esp Disconnect Problems
To: None <port-sparc@NetBSD.ORG, jim.reid@eurocontrol.be>
From: Matthew Jacob <mjacob@feral.com>
List: port-sparc
Date: 10/16/1996 09:51:00
>From owner-port-sparc@NetBSD.ORG Wed Oct 16 01:29:21 1996
>
>You also have to consider how Big Companies manage their software and
>their programmers. One of the metrics they use is the number of bugs
>per X lines of code. The lower the number, the higher quality the code
>supposedly. This gives programmers and their managers an incentive to
>pad out their software so they get bonuses and promotions from the
>beancounters because of the code's is deemed to be high(?) quality....
>

No, it wasn't that many lines of code because of this. It was that
many lines of code because it was done in ~ 3 months (I started on
the 4.0.3c project, if I recall, in March of '88- we held meetings
about the 'new' Sun software architecture and came up with a design
everyone could live with by late May/early June. I implemented all
of the pieces (ncr 53c80 host adapter, middle layers, sd target driver
and st target driver (which, unfortunately, was mostly copied from
the then extant SunOS driver), using a P2 build 4/110 as the testbed
over the next 2-3 months. Did the ESP host adapter driver. I vaguely
recall getting first SS1 boards back from Solectron in late August).
This seemed a lot to get done during a short period of time with other
things to do, so please forgive me if the driver got bulky (whine....)


The trouble with Big Companies, aside from the metrics issues (and
"headcount", etc.) is more that you are rarely allowed to take the
time to do it right at first, and almost never are allowed to go
back and do the most important thing: throw out the first version
and do the second one.