Date: 19 Mar 2002 20:35:08 +0100 From: Dag-Erling Smorgrav <des@ofug.org> To: Quincey Koziol <koziol@ncsa.uiuc.edu> Cc: freebsd-fs@freebsd.org Subject: Re: Filesystem books? Message-ID: <xzp4rjc8i6r.fsf@flood.ping.uio.no> In-Reply-To: <200203190542.g2J5gJi04830@sleipnir.ncsa.uiuc.edu> References: <200203190542.g2J5gJi04830@sleipnir.ncsa.uiuc.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
Quincey Koziol <koziol@ncsa.uiuc.edu> writes: > > This basically looks like strongly typed filesystem - you could build > > this on top of any existing filesystem in FreeBSD using extended > > attributes and a userland library. Storing it in a single XML file is > > IMHO a regression, especially from a performance standpoint. Or did I > > miss some crucial point? > Sorta, the crucial part of the HDF5 library and file format is that the > files and library are designed to be portable between many different types > of machines, ...so you end up with something that sucks equally on all supported platforms. Sorry to sound so negative :) I'm not sure if I got my point across, BTW - what I'm saying is that you can implement HDF (or equivalent functionality, anyway) as a userland library that will run on any system that offers POSIX filesystem semantics and (non-POSIX) extended attributes, instead of reinventing the wheel just to end up with something that will crawl like a crippled snail. It might not look as cool on your CV when you're done (not enough buzzwords) but it'd be a damn sight more useful. The biggest hurdle would be adapting the library to the different EA APIs out there. That being said, HDF makes a neat interchange format, once you get rid of the silly limitations on object size & count. DES -- Dag-Erling Smorgrav - des@ofug.org To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-fs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?xzp4rjc8i6r.fsf>