This is a script for hosting and downloading a static set of files over DNS
without using a custom DNS server. If you give dnsfs.rb
a path to
a directory, it will create BIND-compatible TXT entries for hosting the files
inside. If you give dnsfs.rb
a domain name, it will show you the list of
hosted files and ask which one you want to download.
How fast is it? I downloaded a 568K file from my nameserver in 293 seconds... so about 1981 bytes/second. Of course it depends on the RTT, and it can probably be sped up a lot by sending requests in parallel (this script doesn't).
NOTE: I'm pretty sure something like this has been done before... I just wanted to write it myself and couldn't find the reference. Please let me know of similar things!
Related Work
Note: Intermediate nameservers may cache your requests. Please be respectful and
query the authoratative nameserver directly using the --nameserver
option!
$ ruby dnsfs.rb --download --nameserver dnsfs.defuse.ca dnsfs.defuse.ca
1. manifesto.txt (size: 3880 bytes)
2. longcat.jpg (size: 580516 bytes)
Which one?
1
.....................
File written to /tmp/manifesto.txt.
$ ruby dnsfs.rb --generate test
f1info.dnsfs IN TXT "Name: foo.txt / Size: 31"
f1p1.dnsfs IN TXT "VGhpcyBpcyAKYSBmaWxlCm5hbWVkCmZvby50eHQhCg=="
f1p2.dnsfs IN TXT "~EOF~"
f2info.dnsfs IN TXT "Name: bar.txt / Size: 558"
f2p1.dnsfs IN TXT "VGhpcyBpcyAKYSBmaWxlCm5hbWVkCmJhci50eHQhClRoaXMgaXMgCmEgZmlsZQpuYW1lZApiYXIudHh0IQpUaGlzIGlzIAphIGZpbGUKbmFtZWQKYmFyLnR4dCEKVGhpcyBpcyAKYSBmaWxlCm5hbWVkCmJhci50eHQhClRoaXMgaXMgCmEgZmlsZQpuYW1lZApiYXIudHh0IQpUaGlzIGlzIAphIGZpbGUKbmFtZWQKYmFyLnR4dCEKVGhp"
f2p2.dnsfs IN TXT "cyBpcyAKYSBmaWxlCm5hbWVkCmJhci50eHQhClRoaXMgaXMgCmEgZmlsZQpuYW1lZApiYXIudHh0IQpUaGlzIGlzIAphIGZpbGUKbmFtZWQKYmFyLnR4dCEKVGhpcyBpcyAKYSBmaWxlCm5hbWVkCmJhci50eHQhClRoaXMgaXMgCmEgZmlsZQpuYW1lZApiYXIudHh0IQpUaGlzIGlzIAphIGZpbGUKbmFtZWQKYmFyLnR4dCEKVGhpcyBp"
f2p3.dnsfs IN TXT "cyAKYSBmaWxlCm5hbWVkCmJhci50eHQhClRoaXMgaXMgCmEgZmlsZQpuYW1lZApiYXIudHh0IQpUaGlzIGlzIAphIGZpbGUKbmFtZWQKYmFyLnR4dCEKVGhpcyBpcyAKYSBmaWxlCm5hbWVkCmJhci50eHQhClRoaXMgaXMgCmEgZmlsZQpuYW1lZApiYXIudHh0IQpUaGlzIGlzIAphIGZpbGUKbmFtZWQKYmFyLnR4dCEK"
f2p4.dnsfs IN TXT "~EOF~"
f3info.dnsfs IN TXT "~EOL~"
The naming convention makes it easy to download a file with a standard /bin/sh
command:
for i in `seq 21`; do host -t TXT f1p$i.dnsfs.defuse.ca|cut -d '"' -f 2|base64 --decode; done|less
Alternatively, using dig
:
for i in `seq 21`; do dig f1p$i.dnsfs.defuse.ca TXT|grep TXT|grep '"'|cut -d '"' -f 2|base64 --decode; done|less
All you have to do is replace f1
with the file number you want and seq 21
with the number of parts in the file (the ceiling of the file size divided by
189).
For each file K in 1..N, there are a set of TXT entries:
- fKinfo: File K's name and size.
- fN+1info: End of list marker:
~EOL~
- fKpP: Part P of file K base64-encoded. Last one is
~EOF~
.