1984/5
53. CRITIQUE OF 1984/5
Any critique of 1984/5 must decide whether the project had a reasonable
return on the 10 man years of effort invested during the year.
The achievements in the period were:
-
Hardware: another 60 systems were installed, bringing the
total number to 200. A large proportion of these systems had to be upgraded
to run PNX2.0. This involved a great deal of discussion with users and ICL
to establish a sensible pattern of work.
-
PNX: nearly all the early systems were upgraded to PNX2.0
during the year 1984 or early in 1985. This involved a large amount of
distribution of sets of floppies together with installation at user sites.
-
Maintenance: first line maintenance was provided on 200 systems
throughout the year. This was particularly arduous with the amount of
hardware and software change going on throughout the year.
-
Assessment:
a major and thorough assessment of alternatives to
the PERQ was conducted during the year resulting in SUN being added to the
Common Base.
-
SPY and WW: many additions to Spy were made during the year and
the WW set of window manager utilities were much appreciated by the user
community.
-
GKS:
GKS was made available on the PERQ - the first commercial
implementation of GKS.
-
User Support:
the service desk answered a stream of user queries
throughout the year.
-
Communications: work continued on trying to make the
Cambridge Ring/GPIB interface work at a reasonable performance level
without a great deal of success.
-
Management:
there was continual negotiation with ICL throughout
the year concerning the PNX2.0 upgrade, 16K WCS costs, future of PERQ
developments and the support for PERQ1s beyond PNX2.0. These achieved a
much reduced cost for the 16K WCS boards and the PNXSR release probably gave
the PERQ1s an increase in life of at least
2 years and a significant boost in performance.
There is no evidence to suggest that this level of effort did not produce
good returns throughout the year.
Assessing ICL's performance during the period:
-
Hardware:
there was general
agreement that the PERQ2 was
a significant hardware advance
over the PERQ1 systems.
ICL's
hardware expertise had made significant advances in the
manufacturing, producing a more robust system which was quieter and better
designed.
-
Future Developments:
the PERQ3 had the ability
to be a
significant system in the market place. The continuing lack of a colour
system was a major flaw in the ICL products.
-
UNIX: although taking a year and a half, PNX5 and PNXSR were
eventually reasonable UNIX systems providing good performance over PNX2.0.
The continual flirting with ACCENT because of the Three Rivers involvement
tended to be a digression which cost both effort and a single-mindedness
in direction.
-
Compilers:
the PNX5 compilers show a significant improvement over
the original PNX2.0 systems. The optimising FORTRAN compiler is also a
significant advance over the earlier system.
-
Communications: Ethernet
and Newcastle Connection provide a distributed interactive
computing facility comparable to the
facilities available from ICL's competitors.
-
X/OPEN:
ICL has
a long term commitment
to UNIX and making
European applications software available on the PERQ.
-
Applications Software:
although the applications software
available on the PERQ grew, it was not as fast as its competitors.
This was partly due to non-availability of good compilers, the lack of a
colour system, not sufficient aggression by ICL to entice vendors and the
lack of a US market base.
Assessing SERC's position during the period:
-
Confidence in Programme: lack of confidence in the programme
continued throughout 1984 but was beginning to recede by the end of the
year. The addition of the SUN to the Common Base and the appearance of
better UNIX releases has meant that 1985 has been a relatively good
year with much less adverse criticism. The major problem is still the
fact that no proper local area network
communications has been possible on the PERQ1s.
-
Funding: an increase in funding was provided during the year
1984. However, the great deal of adverse publicity earlier resulted in a
high turnover of staff. As well as disrupting progress, low salaries in
SERC meant it was difficult to recruit the manpower into the project
and the effort in 1983/4 was actually lower than the previous year.
While the effort was allocated, the lack of sensible pay scales caused
difficulties to the project in 1984.
The period 1984/5 has shown significant improvements in the software and
hardware available. The original 256 Kbyte PERQ systems have grown to 1 Mbyte
with microcode enhancements providing significant performance increases over
the last two years.