>>>  1. How even the bulletin could not come up with a graph about the survey
>>> telling us which Starlink applications were important.

Lies, damned lies, and ...  These were preliminary results, but 
potentially dangerous.  There is no weighting or measure of uncertainty. 
How many people voted 5 for HXIS (two I see if the reader bothers with 
the article).  These graphs appear to be merely the raw results without 
any interpretation or almost no caveats.  It's hard to compare a tightly 
focused subroutine library used by perhaps ten people with a big package 
that has to cover the bases of hundreds of differing views.

It would have helped to have done a cricket-averages approach of a 
minimum number of respondents to include in the first three graphs.

You had to study the full analysis published elsewhere (in the SGP 
series) to gain insight.  Looking back you had to be aware of the 
software status not just of the time, but from years before. 
Perceptions last.  I recall one survey where some people were asking for
hypertext versions of SUNs although we had created them a couple of 
years prior to the survey.

> The graphs are
>>> appalling in that the top 10 software packages include only one starlink
>>> product (news). Surely a version could have been done that filtered out
>>> latex and pine etc.

I have a vague recollection that this point was aired at a Project 
meeting.  Most of us weren't on the editorial board.  It's blatantly 
obvious that the general tools for editing, mail, and document writing 
would be near the top of the importance league.  Presumably the Fortran 
compiler was just off the list for building user applications with NAG 
and PGPLOT.  We failed to sell the various infrastructure libraries 
provided to make applications-writing easier over previously established 
brands.  Some people didn't want to be tied to an environment, so 
preferred to write their own equivalents (STARMAN is a prime example).

We know how popular Figaro was, because that's what users knew best and 
the first general package to use hierarchical data structures. 
Anything new would be compare unfavourably with Figaro.  Recall how 
difficult it was to move over some JAC users to CCDPACK and KAPPA.

>>>  2. How much effort was spent on the IRAF interopability (with the vague
>>> hope that IRAF would adopt some of the Starlink low level libraries). Do we
>>> have an idea of how many people actually used the IRAF interface to Starlink
>>> packages?
>>
>> For my stuff, I think I must have two, or maybe three, enquires in
>> total to do with IRAF interoperability. Another example of us spending
>> ages developing clever features that hardly anyone used.
>
> Same for me. A hand full of queries over the years.

I spent a fair bit of time trying to work out the various IRAF formats 
to make IRAF2NDF and NDF2IRAF give the automatic format conversion, I 
had no role in the implementation of applications under the cl.  It was 
a handy proof of concept in terms of adapting Starlink to work with 
FASE, but for practical use, I doubt that we had more than a few users. 
Most were distinct camps and didn't want to mix'n'match systems.

Malcolm