Hi,

In the recent Knesset Internet Committee discussion, Avi Cohen, CEO of
Machba, claimed that the problems of software compatibility with
Microsoft products is insurmountable. Here is the story of one such
"insurmountable" problem, with one of Machba's own system.

Two of the important services offered by Machba are the nation-wide
library catalogs, the Israeli Union List
(http://libnet1.ac.il/~libnet/uli/) and Israeli Union List of Serials
(http://libnet1.ac.il/~libnet/uls/). These sites, based on the ALEPH500
software produced by Ex Libris Ltd. (http://www.aleph.co.il), work only
with Microsoft Internet Explorer.

To quote their information page, "Currently, the ULI web catalog
supports only Microsoft's Internet Explorer� browser, version 5 or
higher" (http://libnet1.ac.il/~libnet/uli/uli01_14-2.htm). Indeed, with
none-MSIE browsers, the Hebrew is messed up.

A little investigation shows that this problem is completely
server-side. Apparently the Aleph server checks the User-Agent HTTP
header, and if it agent is not MSIE then it mangles the page, producing
incorrect data which is obviously misrendered by any browser. The
attached message explains this in detail; the related samples can be
found here:
http://dl.tromer.org/temp/test1.gif
http://dl.tromer.org/temp/test1.html
http://dl.tromer.org/temp/test2.gif
http://dl.tromer.org/temp/test2.html

I sent a detailed technical account of the matter to Aviva Shichor
<[EMAIL PROTECTED]>, Aleph coordinator at Haifa University. (she's
the contact given at the ULS information page). Following an irrelevant
response, I phoned her on 22 May 2002. Her answers were highly
nonprofessional, culminating in her explaining that she "never liked
Unix" and that she doesn't "care what operating system I use -- this
system works with Windows". She refused to hear an explanation of the
technical details, directing me to Ex Libris but refusing to contact
them herself or to suggest whom to speak to. Following that, I sent the
attached message to Elhanan Adler <[EMAIL PROTECTED]>, Director of
MALMAD. It was promptly and courteously acknowledged by Shmuel Moss
<[EMAIL PROTECTED]>, System Librarian, who forwarded it on to Ex
Libris. During the 7 weeks that passed no progress is evident, even
though it's probably just a matter of commenting out an easily located
piece of rogue code.

  From the above, the following is obvious: Ex Libris Ltd. did not make a
reasonable effort to support non-MSIE browsers (a _charitable_
explanation would be that they tried incompetently and messed things
up), and MALMAD did not list support for non-MSIE browsers among the
requirements of the new Aleph system, or did not enforce this requirement.

Specifically, it seems reasonable to conclude that Machba, or at least
MALMAD, does not care about support for non-MSIE browsers. The opinion
of the Machba CEO, as expressed in the Knesset committee, should be seen
in this light.

Additional inquiries concerning the specific technical problem may help
hasten its resolution, so please consider making them.

      Regards,
        Eran Tromer



--- Begin Message ---
Hello,
I would appreciate your help in the following matter.

I am trying to access ULS and ULI using Unix-based browsers that have 
Unicode support (e.g., Mozilla, Netscape6+, Konqueror). I have reached 
some curious conclusions which you may find interesting. The bottom line 
is as follows: these browsers could work perfectly with the new Aleph 
system, and the only reason they do not is that some server-side 
component of the Aleph system sends incorrect data if the client is not 
Internet Explorer. This is a bug in the Aleph system: if it treated 
Mozilla and Internet Explorer *equally*, then both would work.

At the Weizmann Institute there are many Unix users who currently do not 
have a convenient method to use ULS and ULI, because Microsoft Internet 
Explorer is not available on these platforms. I think that a 
country-wide academic service should support all of its users. 
Fortunately, my results show that this can be easily achieved -- all 
that is required is disabling the incorrect browser-specific code in the 
server.

Details follow.

At first I tried using Mozilla normally, and indeed the Hebrew was 
garbled (see attached screenshot: test1.gif). However, looking at the 
actual HTML code sent by the Alpeh server shows that what was sent is 
not valid Unicode: inside every "<x>...</x>" element, where there should 
be Unicode there is garbage (see attached test1.html).

I then tried a little experimentation. In every HTTP web request, the 
browser sends a "User-Agent" string describing the browser version. 
Usually this is used just for informative purposes (logging and 
debugging). None the less, I told my Mozilla to identify itself as 
Interet Explorer by sending the appropriate User-Agent string rather 
than the default one. With this change, everything worked correctly (see 
attached test2.gif). Looking at the sent HTML (test2.html), it's seen to 
be *different* than what was sent before, and indeed contains correct 
Unicode.

It thus seems that the Aleph server examines the User-Agent string, and 
it thinks that the browser is not Internet Explorer then it 
intentionally sends different (and apparently garbled) data.

I also used a command-like utility ('wget' on Linux) to retrieve the 
pages with different User-Agent settings, and got similar results -- the 
server sends correct data when it thinks it is contacted by Internet 
Explorer, and incorrect data otherwise.

Technical details: both screenshots are of Mozilla 1.0RC2 (latest 
version, equivalent to Netscape 7.0 beta) running on on Windows.
I also tested with Mozilla 0.9.5 (roughly equivalent to Netscape 6.0) 
and with Konqueror 3.0.0 -- the conclusions were identical. To change 
the User-Agent settings I used UABar (http://uabar.mozdev.org), which is 
the funny toolbar you see in the screenshot. The URL seen is:
http://aleph.libnet.ac.il/ALEPH/5L6NTLTF5JP1Q7VK2B8S7QURNL331C13M93GPT9RF9LREJV985-07786/full-set/000001-999

For Mozilla, I did the testing both on Windows (Windows 2000 
Professional) and on Unix (RedHat Linux 7.3), and again the conclusions 
were identical. Specifically, Mozilla on Linux indeed dislayed 
everything correctly when the Aleph server was tricked into sending 
proper Unicode.


This issue is troublesome for many (myself included). Since technically 
it is easily corrected, I hope you will be able to get it fixed. Also, 
in general I urge you to consider the convenience of non-Windows users, 
now that there exist many browsers with excellent standard compliance 
and full Hebrew support.

    Sincerely,
      Eran Tromer
--- End Message ---

Attachment: msg20528/pgp00000.pgp
Description: PGP signature

Reply via email to