perlman@wanginst.UUCP (07/18/86)
Here are the raw results from my shar survey. There were many
respondents, many of whom included interesting comments, some of
whom wrote long dissertations of their views of sharing issues. I thank
you all, especially the more opinionated. I am at the end of a summer
course, and don't have time to summarize the many comments, or give my
interpretations of the data. I think I will have time in about a month.
Okay, I have to make some comments. One is that I will be writing yet
another archiver after all; there are several niches unfilled. Please
hold your flames until you see it.
Number of respondents: 107
How many times have you read a shar file?
1-5 6-20 20+
5 19 83
Some people took this to mean "read the contents by looking at
the text inside" but I meant unpacked. Survey design is never
easy; I always end up writing vague questions.
How many shar programs do you have?
3+ 2 1 0
51 22 34 1
Where did you get it?
net.sources mod.sources personal system bundle-(K&P) news
81 71 31 21 3 3
Lots of people write their own shars, and many people have several.
How many shell archives have you built?
0 1-5 6-10 11-20 21+
11 22 18 14 41
I was surprised that some people who have built 1-5 archives
have written their own shar programs.
How have you used these archives?
mail-to-user mail-to-self net-submit storage compression
92 66 53 10 10
This is the most interesting result, to me. The last two categories are
write-ins, so the numbers might have been higher if I had anticipated
the category. I should have, because I use shar for storage too. Some
people said they used archives to reclaim inodes. Some storage users
stressed the need to save files in human reabable format.
A few people said they rated features based on how difficult they
would be to implement. The scales I used did not reflect that
a few people were opposed to certain features, or had specific
ideas about how they should work. Here is a summary of mean ratings.
8.5 pack plain files
7.5 unix universal
6.7 pack directories
6.6 check transmission
6.1 automatically split large files
5.4 check for funny characters
5.3 print table of contents
5.2 guard against overwriting existing files
5.2 guard against trojan horses
4.3 portable to many systems
3.7 pack binary files
3.7 transmit file protections
3.7 allow unpacking specific files
3.5 transmit links between files
2.8 transmit file time
Rate the importance of the following features (0=not 9=very)
Key:
=====feature number-of-respondents average-rating
rating scale values
number of people responding for each value
proportion of users responding for each value
=====pack plain files 93 8.5
0 1 2 3 4 5 6 7 8 9
1 1 0 2 1 2 0 4 0 82
% 01 % 01 0 % 02 % 01 % 02 0 % 04 0 % 88
=====unix universal 92 7.5
0 1 2 3 4 5 6 7 8 9
1 3 3 4 0 1 4 13 16 47
% 01 % 03 % 03 % 04 0 % 01 % 04 % 14 % 17 % 51
People want to be sure that the user on the other end (news or
a specific user on another machine) can unpack.
=====pack directories 93 6.7
0 1 2 3 4 5 6 7 8 9
4 1 0 3 7 16 2 12 20 28
% 04 % 01 0 % 03 % 07 % 17 % 02 % 13 % 21 % 30
Given the importance of this, it was surprising to me that people did
not think transmitting links would be valuable. We use shar to transmit
projects that have files linked across directories. Also, links would
be good for compression.
=====check transmission 92 6.6
0 1 2 3 4 5 6 7 8 9
1 1 3 7 8 12 5 13 16 26
% 01 % 01 % 03 % 07 % 08 % 1 % 05 % 14 % 17 % 28
=====automatically split large files 92 6.1
0 1 2 3 4 5 6 7 8 9
3 1 3 8 4 13 16 15 15 14
% 03 % 01 % 03 % 08 % 04 % 14 % 17 % 16 % 16 % 15
I was surprised to see how high this was rated. Several people wrote
me about a variety of transmission problems with large files/archives.
=====check for funny characters 84 5.4
0 1 2 3 4 5 6 7 8 9
9 4 4 4 8 12 5 12 8 18
% 10 % 04 % 04 % 04 % 09 % 14 % 0 % 14 % 09 % 21
You would not believe how many funny characters there are and what
different systems do with them. It is impossible to maintain human
readible archives and be portable across all networks and systems.
=====print table of contents 91 5.3
0 1 2 3 4 5 6 7 8 9
7 4 2 11 5 10 13 21 13 5
% 07 % 04 % 02 % 12 % 05 % 1 % 14 % 23 % 14 % 05
=====guard against overwriting existing files 90 5.2
0 1 2 3 4 5 6 7 8 9
4 8 4 9 4 21 11 7 7 15
% 04 % 08 % 04 % 1 % 04 % 23 % 12 % 07 % 07 % 16
=====guard against trojan horses 88 5.2
0 1 2 3 4 5 6 7 8 9
10 4 7 6 4 14 6 11 7 19
% 11 % 04 % 0 % 06 % 04 % 15 % 06 % 12 % 08 % 21
Many people said this was impossible. Even if the files can be unpacked
safely, there may be a program with a bomb in it.
=====portable to many systems 92 4.3
0 1 2 3 4 5 6 7 8 9
21 6 4 7 5 16 8 5 6 14
% 22 % 06 % 04 % 07 % 05 % 17 % 08 % 05 % 06 % 15
Bimodal. There are people who work on several systems, and those who don't.
=====pack binary files 90 3.7
0 1 2 3 4 5 6 7 8 9
19 6 14 10 7 8 5 6 6 9
% 21 % 06 % 15 % 11 % 07 % 08 % 05 % 06 % 06 % 1
The low rating here surprised me, even though I have never sent binaries.
=====transmit file protections 92 3.7
0 1 2 3 4 5 6 7 8 9
18 10 6 13 2 14 9 13 3 4
% 19 % 10 % 06 % 14 % 02 % 15 % 09 % 14 % 03 % 04
=====allow unpacking specific files 91 3.7
0 1 2 3 4 5 6 7 8 9
15 6 12 13 12 11 6 6 4 6
% 16 % 06 % 13 % 14 % 13 % 12 % 06 % 06 % 04 % 06
=====transmit links between files 89 3.5
0 1 2 3 4 5 6 7 8 9
17 7 10 15 9 8 8 7 5 3
% 19 % 07 % 11 % 16 % 10 % 0 % 0 % 07 % 05 % 03
=====transmit file time 92 2.8
0 1 2 3 4 5 6 7 8 9
31 9 6 13 6 8 6 9 2 2
% 33 % 09 % 06 % 14 % 06 % 08 % 06 % 09 % 02 % 02
Except for "make" dependencies, I can't think of a reason for this.
Maybe it would allow using the date as a version identifier.
--
Gary Perlman Wang Institute Tyngsboro, MA 01879 (617) 649-9731
UUCP: decvax!wanginst!perlman CSNET: perlman@wanginst