User: Password:
|
|
Subscribe / Log in / New account

A kernel message catalog

A kernel message catalog

Posted Aug 8, 2008 20:47 UTC (Fri) by giraffedata (subscriber, #1954)
Parent article: A kernel message catalog

we are dealing with companies like IBM which have been creating massive binders full of kernel message documentation for several decades.

Do they still? That practice began when computer storage was too expensive to keep all that information online. Today, a message manual increases cost for everyone involved. And that's true even if the manual is stored online.

It irks me endlessly that software engineers seem to think they have to pay by the word for error messages, and that it's somehow undignified to come right out and tell the user what's going on. Here we have a proposal to send the user on a treasure hunt to find out why his module wouldn't load. The whole paragraph explaining the XPRAM devices problem could be right in the printk, and the second paragraph giving advice about it would not be unappreciated, right there in the log, either.

I understand the laziness that makes people not compose error messages, but in this case, the developer does write the message; he just sequesters it.

There are some messages that happen so normally that a full explanation would be an inappropriate use of message space, but the example isn't one of those.


(Log in to post comments)

A kernel message catalog

Posted Aug 9, 2008 1:05 UTC (Sat) by erwbgy (subscriber, #4104) [Link]

Embedding the text directly in the source files would make translations more difficult, which
would probably mean they didn't happen, so I don't think this is a good idea.

Instead, perhaps some post-processing of the source code could be done to embed the text for
your chosen language, so you do get the text directly in the log, as you suggest.

A kernel message catalog

Posted Aug 9, 2008 10:07 UTC (Sat) by dark (guest, #8483) [Link]

Embedding the text in the source code needn't make it harder to translate -- gettext solves exactly this problem and I'm sure the technique could be adapted for the kernel. The drawback is that if the kernel outputs translated messages directly then the English-language version is not available when submitting bug reports or asking kernel developers for help. For that, a unique identifier would help.

A kernel message catalog

Posted Aug 9, 2008 16:31 UTC (Sat) by giraffedata (subscriber, #1954) [Link]

The drawback is that if the kernel outputs translated messages directly then the English-language version is not available when submitting bug reports or asking kernel developers for help. For that, a unique identifier would help.

Note that the current proposal works the other way around: the kernel produces English and a unique identifier, then the user looks up the identifier separately and finds a more complete explanation possibly in another language. And in mine, the difference is that the English part is longer and many users don't need the additional step.

"harder to translate" might mean harder for someone to generate the other-language message repository, as compared to where the English version is in a separate file of message text. But in the current proposal, I believe the full English text is in the source file anyway; it's just in comments instead of in the actual message. A program extracts this text from throughout the kernel and a translator uses that extract to generate a message repository.

A kernel message catalog

Posted Aug 11, 2008 1:35 UTC (Mon) by ajf (subscriber, #10844) [Link]

They still rigorously define the error codes - more out of habit than a desire to be helpful,
I suspect, since almost all of the ones I've encountered since I first encountered WebSphere
Application Server seem to be documented as "Something didn't work. Don't do that.".


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds