As I was expecting, some assholes started throwing around snide remarks about my inquiries yesterday. I actually embarked on a farily large rant concerning the way people are simply not mature enough to accept that I am just trying to help out a few friends (and keep trying to figure out whether myself or my company were having difficulties due to some sort of psychiatric need for compensation), but I canned it and just posted a short(ish) disclaimer.
Think what you want, say what you want - you're not hiring any of my friends that way. And if I ever want to change jobs, I'm not even going to give you a chance of hiring me.
You're simply not worthy.
Tuning with /proc
Solaris was the first OS I came across where I did (minimal) kernel fine-tuning at runtime. Linux also lets you do some changes without rebooting the system, and this article has a nice overview of the stuff available under /proc to do that (I use /proc all the time for collecting performance counters, but some of this was new to me).
Too Much Storage, Too Many Places
Yes, you can have too much storage. The main issue I have with data is replication without needless waste, since the 10-20GB of data I can actually call my own (actual documentation, code and media I've created) is taking up around 130GB of disk space at several locations under the guise of various backups, snapshots and replicas. Of course there are other issues: Searching, remote access, backups, security, multi-platform support, etc.
For instance, my Wintel laptop (where I have all my work e-mail - 2GB worth of it - my project documentation and references - 4 year's worth, another 5-7GB - and around 1GB sources, build trees, etc.) is partially replicated to an office share, our CVS repository, our Exchange server, our development server, and (weekly) to a Firewire HD at home. All of it synced using different protocols and tools (ssync, rsync, straight SMB, etc.), mostly by hand. My iBook doesn't store e-mail (I aggregate everything on an IMAP server at home) or code (I have my own CVS too), and Apple's updates are diligently downloaded by hand and filed away on another box. And my Linux box... Well, that's just too hairy.
What I need is a filesystem that:
- Is accessible from anywhere in a reasonably secure fashion
- Works properly over low-bandwidth links (SMB over any sort of VPN performs hideously)
- Has minimal support for all my platforms (I can live without Mac resource forks, but Linux and Windows boxes have to read the rest)
- Provides a coherent namespace across all my boxes and disk volumes
- Performs transparent replication of critical data (I should just set an attribute on a directory and be certain it's stored on more than one box)
- Supports some sort of "offline" or "disconnected" operation (caching, directory indexes only, whatever)
Volume abstraction (not having to care about specific volume sizes and splitting files across volumes) would also be great, but I'm being realistic here - I'm not asking for Star Trek-like filing systems... Yet.
So once in a while I get entirely fed up with it all and start searching for a better way to do things. However, there seems to be fairly little progress on the distributed filesystem field. Over the past 10 years or so, I've been in places where people used SMB, NFS, AFS, Coda, Microsoft's DFS, AppleShare, the works. But, as usual, there is no single solution that addresses all my needs. After seeing this discussion, I'm now revisiting stuff like Unison and OpenAFS, but nothing seems good enough.
The PEG-NX70V In Hand
rage got his hands on a PEG-NX70V, and the pics are in the photo album. The machine is easily the best Palm device I've handled to this day (a bit big, but the Sony industrial design and usability put most other Palm and Pocket PC devices to shame).