Users!

[andygates]

Which of these did I actually send tonight at 10pm after checking in on my server migration (after a great night-school session: vertical down outside corner welds, you are my bitches)?

1: "Dear users, the migration of your data to the new, spacious and high-performance server cluster will not go ahead tonight as planned due to technical issues which we'll investigate as soon as we can.  In the meantime, please continue to work on the current server as before."

2: "Dear users, the migration of your data from the creaky old fossil box to the preposterously over-spec server cluster will not go ahead tonight as planned because you've chosen to use such staggeringly long filenames that they crashed my migration routine.  I mean, come on, I'm all for lucid file structures but whole sentences as folder names?  You've saved every webpage you've ever found interesting with its full name, then given a "that huge name - comments" subfolder with more inside that, ad infinitum?  Easy, tiger.  We gave you a namespace but you eated it.  Your linguistic legerdemain did things that Windows, in its deep stupidity, permits but cannot handle, which I shall try to cheat my way around as soon as I can face it.  In the meantime, please continue to create novella-size file structures on the current server as before."

Comments

Re: Here's the thing

[andygates.livejournal.com]

You are *absolutely* correct and thank you for bringing me to task on that. My real vexation is with Windows for allowing this situation to occur in the first place. It's 2008! We have flying robot killer drones and LED contact lenses and laptops that fit inside envelopes. The failure to handle an arbitrary namespace for user file names is inexcusable. Especially when we encourage verbose naming to provide context.

There are two strands of stupidity that manifest: first, the user network drive X:\myfiles has a length of 10 characters; S:\userdata\shares\team1\myfiles, the share's root on the server, has a length of 32 characters. It's amazing the number of times that makes a difference and the check isn't performed against the volume root. Second, I can rename X:\myfiles\foo\bar to X:\myfiles\foo_in_modern_society_-_the_value_of_a_placeholder_in_semantic_interchange\bar and it doesn't check downstream to see if it has made anything illegally long.

They're neither trivial to fix, but they are trivial to work around: allow arbitrary large filenames.

Bloody Windows.

Re: Here's the thing

[gedhrel.livejournal.com]

Heh, been bitten by this before* :-) Might find that UNC pathnames and/or the mount/join** DFS stuff will let you set up something that xcopy will work on.

* In about 1995. Even then it was abysmal since remote filesystem protocols a decade older would cope with it.

** can't remember what they call that operation.

Re: Here's the thing

[andygates.livejournal.com]

Shameful, isn't it? Ho hum.