[Freecats-Dev] RB Manager - first comments after a quick review

[Henri Chorand]

Simos,

> http://oss.software.ibm.com/icu4j/demo_tools/RBManager.html
>
> I run it and I saw the initials 'TMX' somewhere in the 'Import' section.
> Of course I did not manage to make any good use of it. If someone can
> give a short summary, would be appreciated.

Thanks for bringing this to our attention, we definitely must take a 
little time to assess it.

I noticed that some of the menu links on the left are dead, but most of 
them are most probably the ones included in page:
http://oss.software.ibm.com/cvs/icu4j/~checkout~/unicodetools/com/ibm/rbm/docs/index.html

The limited amount of information given here only enables me to provide 
the following feedback. Here are my comments:

They provide a rather common-sense view on localization - nothing 
revolutionary here.

Around 8 years ago or so, IBM launched the very first CAT tool on the 
market, IBM Translation Manager, with which we made a few projects at 
Kemper DOC for indirect customers, via large localization agencies 
(including a storage appliance for IBM). We did not like working with 
it, because we could not quite achieve the same levels of productivity 
(while not being able to raise our prices accordingly). A lot in the 
interface was in character mode, and the learning curve (in order to 
become able to use it fast enough) reputedly represents a few months.

At that time, it was a proprietary (and expensive) software. Trados came 
into light a bit later on and took off rather quickly because of its 
quite better user-friendliness and large set of features. I did not find 
any reference to IBM Translation Manager here.

Even if IBM did make their Translation Manager open source, I would not 
want to use their code as a basis on which to start our project. Of 
course, they might have done a quite valuable work in terms of algorithm 
logic, it's their interface which is a mess.

RB Manager seems purely Java resource file-oriented. It seems it was 
created in order to address issues rather specific to this environment - 
the sometimes huge number of places where to search for localizable 
strings in a Java project - and to ensure consistency between the 
translations of the same strings when found in a large number of project 
files. When you read their tutorial, you find that an important portion 
of the interface is dedicated to grouping resource files, the 
translation bit is only a pair of text fields for source & target 
strings. Mind you, if I was heading software localization at IBM, maybe 
I would begin by teaching developers about the virtues of grouping 
translatable resources in the smallest possible number of dedicated 
files, instead of spreading them everywhere within a project's source 
code ;-)

They don't speak about translation memories. Therefore, I believe this 
program can only retrieve perfect matches, not fuzzy ones. This stuff 
also clearly looks like a tool designed by a developer and intended for 
other developers. I would say it's an interesting Java-only prototype. 
Also note that this software seems very restricted in scope and already 
asks for 64 Mo RAM free in order to run. Well, it's Java...

One of the nice things seem to be:
"Import and export support for TMX formatted XML translation files"
TMX format is a simple one, which we translators regularly use. I'll 
generate one and post an excerpt when back in front of my office computer.

Both our (or any, for that matter) TM database and bilingual working 
formats represent a bigger issue here.

So, all in all, I guess we could contact RB Manager's chief developer, 
Jared Jackson (address@hidden), in order to try to interest IBM 
in our Free CATS project, which I believe has much more potential.

Once we have set up our bilingual document working format, they might be 
wanting to provide nice conversion filters from and to Java source files 
- or much more help if they want to help us defeat Micro$oft-Trados.

Last... I did not try to run this software, as I don't presently have 
Java resource files at hand. I suppose I could try to download the 
source code of any Java free software stuff, but I would also need to 
check first if it does contain translatable text resources. Maybe 
Bertrand (who knows Java) can do this some time this week.

So, I'm afraid this is all I can say about it for now.


Regards,

Henri