| top | introduction | details | conclusions |
The information on the browser is collected automatically by a combination of javascript code and a CGI script, whenever the user connects to a web page like this (the particular example is inactive, i.e. mostly harmless).
The code may not be perfect, but is reasonable, and is available to anybody interested for modification and re-use.
In half an hour I collected already about 60 valid hits, and in less than one day about 260 valid hits. At the time the survey was stopped at 11:30 on 14 March, 308 valid hits were collected.
This has to be considered a rather good return, considering e.g. that a more formal survey on astronomical software collected just 130 replies in a longer period of time, and that was considered a fairly good return.
To avoid spurious hits by outside people (or robots) I restricted the access to the survey page to known domains belonging to INAF Structures. This means some "denied access" errors were generated by people which were abroad, or in other Italian institutes, or at home, or on a machine without reverse resolution.
These accesses can be counted, but do not enter the tally of the results since they
were not recorded.
Namely we had 14 invalid hits from an INAF LAN without reverse resolution, 10 hits
from a proxy on a CNR domain, 12 hits from miscellaneous italian domains (university,
INFN etc.), 5 hits from abroad and 3 hits from home.
Valid hits were ingested in a database, which could automatically remove duplicated entries. A duplication occurs when a hit occurs by the same IP address and hostname, from the same browser (name and code name and version).
Users were requested to hit the survey from their standard workstation using their
preferred browser in its default configuration. There was however no way to enforce
it. The above provision should get rid of multiple hits by mistake, but will not get
rid of willing multiple hits using different browsers.
Also in some cases there are systematically a large number of hits from the same
machine, which make me suspect this machine is acting as a proxy server.
Actual results of possible general use or interest follow here below, while those of likely interest only internally are reported separately. The original file (inclusive of duplicated hits) will be also available there.
Lucio Chiappetti INAF IASF Milano
15 Mar 05 17:39
However I noticed that, even with such criterion, there were 7 hosts with multiple
hits. Three of them had respectively 4, 6 and 10 hits, and are most likely proxy
servers. The other have two hits. In most cases the double hits are from a Linux and
a Windows system, so they could be dual-boot systems, but I consider it unlikely.
In one case the double hit was from Linux and OSF.
I also verified that most multiple hits have different screen sizes, which
reinforces the idea of different machines viewed via a common proxy server.
Therefore I waived the rule of indexing on the combination of items listed above,
and decided to ingest all 308 entries as if they were independent.
All results described below refer to this full sample
Of the 308 hits, 43 are therefore multiple hits apparently from the same machine.
Of these 15 have more than 2 hits, and just 5 more than 3 hits (namely 5, 12, 13, 13
and 34), which definitely qualify as proxies. The latter 4 comprise the totality
of the hits from the relevant seat (respectively Palermo IASF and Observatory, Torino
and Padova).
| Platform | count |
|---|---|
| MacPPC | 2 |
| Win32 | 128 |
| Linux ... | 147 |
| Sun | 18 |
| Alpha | 10 |
| HP | 3 |
Notes
Most (132) of the Linux system identify themselves as "Linux i686" , the other with some variant, one as just "Linux" and 3 as "LinuxELF2.n"
The Sun systems identify as "SunOS5.5.1" (10) or "SunOS sun4u" (8).
The Alpha systems identify as "OSF1 alpha" or "OSF1V4" (plus one not identified by well known, i.e. my own old machine !)
The navigator.appName property is not sufficient to tell the actual type of browser. In fact in our database there are only three possible values :
| Application name | count |
|---|---|
| Netscape | 215 |
| Microsoft Internet Explorer | 83 |
| Konqueror | 10 |
The navigator.appCodeName is even less useful. All browsers in our database identify themselves as Mozilla !
The navigator.appVersion property is discussed in some detail in the next
section, but is of little use here. Its format is not very predictable. It usually
has a form like
n.m [something] (something else ; something more ; ...)
where the something in round bracket (always present) may contain indication
of the operating system, the language, the window system, the type of browser or
a further version number). The only case in which it contains a browser identification
is Konqueror, which is already identified at navigator name level !
The navigator.userAgent property is also of little use. It has a form like
Mozilla/n.m [something] (something else ; something more ; ...)
i.e. almost replicates the version string. Exceptionally there is a something
not in brackets before the round brackets, and this is the navigator identification
for Galeon.
Also note that I have truncated all property strings to 50 characters while storing them, so any info provided in longer strings has been lost.
I finally resorted to a semi-manual re-classification using the various pieces of information. This way I was able to separate MS Internet Explorer and Opera, Galeon and Konqueror. I flagged as "Netscape" all stuff which is not IE and has a version 4.x, and as Mozilla everything which has a version 5.0 and is not otherwise flagged. I am still unable to flag Mozilla from Firefox (here one has rv:1.7.2 and the other rv:1.7.3, but I've no idea if that has any meaning).
| Browser | on | count | Comments |
|---|---|---|---|
| Mozilla | MacPPC | 2 | very limited use of Mac |
| Netscape | Win32 | 3 | IE dominant on Windows, but "Mozilla" widespread |
| Mozilla | Win32 | 43 | |
| IE | Win32 | 82 | |
| Galeon | Linux | 1 | non-netscape browsers rare on Linux |
| Opera | Linux | 1 | |
| Konqueror | Linux | 10 | |
| Netscape | HP-UX | 3 | "real" netscape used on vestigial non-Linux Unix, and on Linux ELF whatever it is |
| Netscape | LinuxELF | 3 | |
| Netscape | Alpha | 5 | |
| Netscape | Sun | 10 | |
| Mozilla | alpha | 5 | dominant on Linux (diffuse on Windows, see above) and present also on vestigial Unixes |
| Mozilla | Sun | 8 | |
| Mozilla | Linux | 132 |
| Version | count | notes | |
|---|---|---|---|
| 3.0 | 1 | this is Netscape on Alpha (guess who !) | |
| 4.0 | 83 | this is IE on Win (or Opera on Linux in 1 case) | |
| 4.08 | 1 | these are presumably all versions of Netscape | Linux ELF |
| 4.5 | 1 | Sun | |
| 4.61 | 1 | Linux ELF | |
| 4.7 | 4 | Alpha, Sun, 1 Win98 | |
| 4.76 | 1 | Sun | |
| 4.77 | 6 | Sun, 1 Win95 | |
| 4.78 | 4 | Alpha, Sun | |
| 4.79 | 1 | Linux ELF | |
| 4.8 | 4 | HP-UX, 1 Win98 | |
| 5.0 | 201 | this includes Mozilla, Firefox, Konqueror, Galeon, the Mac's and 43 Windows | |
I resorted to the manual classification described above, where I took the n.m version for what is flagged Netscape, while I extract the version from the "User Agent" string for Internet Explorer, Konqueror, Galeon and Mozilla. I cannot assign a version to Opera, the Mac browsers and Mozilla on Windows
| Browser | internal version | count | comments |
|---|---|---|---|
| Galeon | 1.2.7 | 1 | marginal |
| IE | 5.0 | 3 | |
| IE | 5.5 | 3 | |
| IE | 6.0 | 76 | dominant version among IE |
| Konqueror | 3.1 | 8 | marginal |
| Konqueror | 3.2 | 1 | |
| Konqueror | 3.3 | 1 | |
| Mozilla | 0.9.9 | 4 | total of 34 "low" versions |
| Mozilla | 1.0 | 5 | |
| Mozilla | 1.0.1 | 4 | |
| Mozilla | 1.0.2 | 10 | |
| Mozilla | 1.2 | 2 | |
| Mozilla | 1.2.1 | 7 | |
| Mozilla | 1.3 | 1 | |
| Mozilla | 1.3.1 | 1 | |
| Mozilla | 1.4 | 40 | most diffuse Mozilla "rv" version |
| Mozilla | 1.4.1 | 3 | total of 38 "intermediate" versions |
| Mozilla | 1.4.2 | 6 | |
| Mozilla | 1.4.3 | 9 | |
| Mozilla | 1.5 | 3 | |
| Mozilla | 1.6 | 17 | |
| Mozilla | 1.7.2 | 9 | total of 33 "high" versions |
| Mozilla | 1.7.3 | 14 | |
| Mozilla | 1.7.5 | 8 | |
| Mozilla | 1.7.6 | 2 | |
| Mozilla | n/a | 45 | Only Mac and Windows |
| Netscape | 3.0 | 1 | vestigial, see previous table, heterogeneous versions |
| Netscape | 4.08 | 1 | |
| Netscape | 4.5 | 1 | |
| Netscape | 4.61 | 1 | |
| Netscape | 4.7 | 4 | |
| Netscape | 4.76 | 1 | |
| Netscape | 4.77 | 6 | |
| Netscape | 4.78 | 4 | |
| Netscape | 4.79 | 1 | |
| Netscape | 4.8 | 4 | |
| Opera | n/a | 1 | marginal |
| Screen size | count | Windows | Unix |
|---|---|---|---|
| 1280x1024 | 193 | 72 | 121 |
| 1024x768 | 72 | 38 | 34 |
| 1400x1050 | 10 | 6 | 4 |
| 1152x900 | 8 | 8 (Sun) | |
| 1600x1200 | 6 | 6 | |
| 1152x864 | 5 | 5 | |
| 1280x800 | 4 | 4 | |
| 1024x1280 | 3 | 3 | |
| 800x600 | 2 | 1 | 1 |
| 1280x960 | 2 | 2 | |
| 1280x967 | 1 | 1 | |
| 1600x1024 | 1 | ||
| 1920x1440 | 1 | 1 |
The screen size can be used when designing a web application to consider the maximum size of a window one can reach (also considering that javascript can resize the window)
Therefore it is only useful to get the window sizes in ranges.
| width | count | height | count | |
|---|---|---|---|---|
| less than 700 | 8 | less than 400 | 6 | |
| between 700 and 800 | 15 | between 400 and 500 | 12 | |
| between 800 and 900 | 58 | between 500 and 600 | 69 | |
| between 900 and 1000 | 53 | between 600 and 700 | 42 | |
| between 1000 and 1100 | 82 | between 700 and 800 | 64 | |
| between 1100 and 1200 | 24 | between 800 and 900 | 108 | |
| between 1200 and 1300 | 65 | between 900 and 1000 | 4 | |
| above 1300 | 3 | above 1000 | 3 |
Apparently most people (contrary to me !! I use 708x632) like to keep their browser window quite
large, or nearly full screen. I instead prefer to occupy only a small portion of the screen,
with a portrait-like orientation (includng browser buttons etc.).
An alternate view can be obtained comparing the window size with the screen size.
| width as % of screen width | count | height as % of screen height | count | area as % of screen area | count |
|---|---|---|---|---|---|
| less than 50% | 4 | less than 50% | 9 | less than 50% | 45 |
| between 50 and 70% | 48 | between 50% and 70% | 62 | between 50% and 70% | 162 |
| between 70 and 90% | 134 | between 70% and 90% | 237 | between 70% and 90% | 103 |
| between 90 and 100% | 89 | ||||
| exactly 100% | 33 |
For what concerns Java we use the result of the navigator.javaEnabled() method. This might not be a definitive answer (some people told me they have Java enabled as a plugin).
Anyhow 231 machines respond to be Java enabled.
Of the 77 machines with Java disabled, 5 are windows, 1 is a Mac and the rest
is Linux.
This information could be used when designing web applications, although the
good old viewable with any browser
principle remains a good design idea.
| top | introduction | details | conclusions |