Flippantly Anticipated Questions

Q: It takes too long to reconstruct a backup after a disk failure.
A: In this instance, you should lobby the legislative bodies in your country of origin to pass ineffective spam-control laws; you might wish to suggest to your local parlimentarian that they model their legislation on the American CAN-SPAM act. Such suggestions are often strengthened by the applications of a "business case" argument, such as "my company's backup strategy, on which my crucial customer data depends, relies on a steady flow of unsolicited mail to harvested addresses. Surely you can see that the needs of today's business should play a determinative role in legislating the Internet." If that fails, try formulating some sort of argument based around copyright.

Q: I need to make multiple backups.
A: In the simplest case, you should simply register a new domain each time you wish to start a new backup, and use that. If your MTA supports subaddressing, e.g. as with Exim and Qmail, use a suitable address prepending (see commandline help for the -p switch). You might also contact your local bulk-email enterpreneur and ask to purchase secondhand .biz domains rendered useless by SURBL. Just imagine the fun you'll have wearing your BOFH hat and announcing to lusers, "Ah yes, your backup. It's on the onlinepillz4u.biz tape. I'll have it restored next month."

Q: Spammers don't seem to be keeping up their end of the bargain -- I always end up with holes in my backup files.
A: Consider the use of a parity generator (parchive or par2). This can reduce the portion of your backup you must have returned to you before you have enough of the data to restore.

Q: Using email for backups seems like it would use a lot of bandwidth.
A: You don't need to handle both sides of a full duplex transaction. Once you've made available the backup addresses, you can purchase copies of your backups on CD from many reputable vendors (such offers will often be provided via email). However, you may find that these CDs are often contaminated with other sites' backups as well -- remember to grep for your own domains, and consider using the -p option for ease of identification.

Q: How do I know when the entire backup has been restored?
A: Consider prepending or appending an MD5 digest of the backup before encoding it with mj. When the file matches the checksum, it's all there.

Q: The encoding scheme doesn't seem very efficient.
A: It uses two address bytes per input byte. By coincidence, the inverse of this ratio is approximately what you can attain through a typical compression algorithm. Taken together, this scheme is a model of efficiency -- or at least identity.

Q: Yeah, but couldn't you have used Base-64 or something?
A: There are some lengths to which I'll go for the sake of a joke. Concocting an RFC822-clean base64 encoder isn't one of them.

Q: You can't be serious.
A: I'm not. You can laugh now.


Devin Carraway <mayonnaisejar @ devin.com>
$Id: faq.html 392 2004-09-29 09:07:06Z aqua $