Greg A. Woods wrote:
[ On Monday, July 2, 2001 at 20:59:49 (-0500), Marius Strom wrote: ]
> Subject: Re: Getting an AS and /18
> c) Take all that spreadsheet data, and throw it into an rwhois server.
> For a /20 worth of addresses, this took me the better part of 3-4 days
> of data entry. I would expect it would take one person 12-16 days for a
> /18 worth of addresses. Need it faster? Throw more people at data
This "spreadsheet" and "data-entry" stuff seems like the wrong way to go
Using a spreadsheet to split out subnet lists seems like way-over-kill.
Don't you already have routing tables or other configuration data that
lists all this stuff somewhere?
As for needing to do data-entry, well all the data comes from computers
in the first place. It should only take a competent programmer a day or
two at most, and maybe even just an hour or two at best, to hack together
something to manipulate it from one form to another as required.
This, of course, assumes that such data is:
B) In a machine-parseable format
If they're not already running rWhois, and think this is a big project,
then chances are one (usually, both) of the above isn't true. It certainly
hasn't been, in most places that I've been which didn't run a database to
track it already.
Humans are amazingly good at drawing useful information from poor formats,
and at noticing errors when testing said information and finding probable
error points (well, OK, humans with any experience in doing so). They're
also amazingly bad, in my experience, about remembering to fix the bad
data, or deciding on a consistant format among themselves, without some
external force requiring it (usually beyond the mere "policy" and into
the "proactive enforcement" level).
On the other hand, using said data entry clerks to do the entry into a
new system that *does* have those features, and test/cross-check/debug
the data as it's being input (as well as cross-checking each other, very
important)... well, error rates are still non-zero, but they do go down