Transcript Slide 1

Contact Data in the RIPE Database
Shane Kerr
RIPE NCC
<[email protected]>
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
1
http://www.ripe.net
Background & Goal
• Certain kinds of data have caused problems
– Domain objects (heavy use by ccTLD’s)
– Person objects (heavy use by ccTLD’s, etc.)
• Cleanups have been made in the past
– Consistency fixes
– Deletions of unnecessary data, one-time and ongoing
• Small numbers of “inconsistencies” not a problem
• Perform some measurement of data quality
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
2
http://www.ripe.net
Contact Data
• Contacts are:
– Referenced by resources recorded in the Database
– Administrative or technical
• Contacts have:
–
–
–
–
Name
Postal address
Phone number
E-mail address
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
3
http://www.ripe.net
Focus on e-mail
• Name impossible to check
• Postal address/phone number difficult to check
• E-mail possible
– Sadly optional for person objects
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
4
http://www.ripe.net
Checking the addresses
1.
2.
3.
4.
Unique e-mail extracted (about 500,000 in all)
Syntax check to remove garbage and bad TLD
Unique domains extracted (about 280,000 in all)
DNS checked
•
•
Algorithm from RFC 2821
MX lookups, with fallback to A lookups
5. SMTP checked
•
•
•
VRFY unreliable
Use RSET, MAIL, RCPT for each e-mail
Minimise connections (only 140,000 unique IP’s)
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
5
http://www.ripe.net
E-mail results
0%
7%
3%
10%
Bad Syntax
Bogus TLD
No DNS Entry
No Server
Server Refused
Okay
80%
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
6
http://www.ripe.net
“Refusal” Codes
550: mailbox unavailable
553: mailbox name not allowed
2%
73%
1%
452: insufficient system storage
1%
451: local error in processing
4%
554: transaction failed
4%
421: service not available
6%
551: user not local
9%
Shane Kerr
all others
.
RIPE 45, May 2003, Barcelona
.
7
http://www.ripe.net
Interpreting the Results
• 20% of e-mail addresses can never be reached
• 80% may still fail
– Depends on mail software and configuration
– Impossible to check further without delivering mail
– Even delivered mail may never be read
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
8
http://www.ripe.net
aut-num results
1%
6% 0%
RIPE, reachable
non-RIPE, reachable
RIPE, unreachable
non-RIPE, unreachable
93%
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
9
http://www.ripe.net
inetnum results
objects
• A significant percentage of
inetnum objects have no 23%
valid e-mail address.
• A much smaller
percentage of actual IP
6%
addresses has no valid email address, but still a
significant amount.
8%
• Most of these are because 4%
the “e-mail:” attribute is
optional in the person
object.
Shane Kerr
.
RIPE 45, May 2003, Barcelona
71%
reachable
non-reachable
no e-mail
88%
IP addresses
.
10
http://www.ripe.net
Conclusions & Questions
• Many networks have no reachable contacts
• “e-mail:” being optional is a significant reason
• Is this a problem? If so, how big of a problem?
• Possible actions:
–
–
–
–
Make “e-mail:” mandatory
Check e-mail reachability on person creation/update
Put a “remark:” on networks with unreachable contacts
Return parent networks if contacts unreachable
Shane Kerr
.
RIPE 45, May 2003, Barcelona
.
11
http://www.ripe.net