db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jan Hlavatý <hla...@code.cz>
Subject Re: java1.4.2 "rws" mode fix: LogToFile.java diff
Date Sun, 05 Sep 2004 14:54:48 GMT
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Daniel John Debrunner wrote:

> Cloudscape has always been written against
> the Java specs, not making any assumptions on how various Java
> implementations implement the spec. I would hope Derby would follow this
> idea.

Even Sun has to work around bugs. Telling users its not our fault will not help them.
For example, the fact that log forcing does not currently seem to work correctly on OSX
JDK 1.5rc1 and thus users may have higher probablility of data loss in crash
when running Derby on OSX is something they will really want to know in advance ;-)
Reliability is key in database systems.

I can see two ways to handle known bugs:

1. Automatic way: check for exact JVM version that is known to have the bug,
then apply workaround when that version of JVM is found. Assume unknown JVM does not have
the bug.
This is the way Sun recommends for working around bugs in JDK.
This could be handled by same mechanisms that configure Derby for different JDK versions.
Or it could even be handled as part of installation process, so it wouldnt have impact
on runtime performance.
Of course, only some bugs can be worked around - if bugs are too serious
and impossible to workaround, Derby should not run and issue appropriate warning, instead
of
failing because of the bug.
That way, users wont be bitten by known bugs, and will be happy.
This would also be good solution for environments where users are not free to pick
the JVM they use.
This is ideal solution for users, and bad solution for developers ;-)

2. Known bugs list: This would require a list of JVM versions with known bugs that
affect Derby, possibly on a web page, or a section in readme file.
Users would have to be directed to it to check their environment.
This requires no code in Derby, it's just and administrative measure, and thus an easy one
;-)

Of course, there may be other bugs than JVM bugs, which would be nice to be aware of too.

Jan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iQEVAwUBQTsot3FDePgyse5HAQJ7FQf/eMpm3dPCCqWE6o3RsOAn9rhBabw7rA10
W88HNjIfLMShaX9DX4YOUFLF+Hyit+VEyqaPH9NE+zOsXQe5EjNQ+DS79W8zQdB7
aG9TdlpedpwTshIddvEjr9NRYg8a0r5jhZUEccvqPO1km953PW5XWAeooi+lst6E
rmfi6FTyRKMZNib/JW4X+nzTjhVL5+PAJGVL3XRA0ApuS71sKn4lSH1RtGFoYfec
aBjOfOxAEY23HcLmJxfzOoxX8KcOEtXIAJatLYF5F0lWBI2POzAGfNoaN5b6SdBT
PglWFkfpeMSs+0z0NwQ23ybFWqxKeXS4LYEOdDlG9u9FS6xzMKim0w==
=Pkv5
-----END PGP SIGNATURE-----

Mime
View raw message