## Burrow the Burrows ##
__Statistics_on_the_subgraph_of_type-1_selectors__
(last update: 20180827 17:00:00 GMT)
### N.B.: THE CRAWL IS NOT COMPLETE, SO THESE
RESULTS ARE PARTIAL AND NOT CONCLUSIVE.
The graph of type-1 selectors is obtained by
associating a node to each menu that is traversed
by `burrow`, and by drawing a directed edge from
menu A to menu B if there exists a selector to
menu B in menu A. As a result, this graph contains
only menus in servers that have been found active
at the time of the crawl.
The subgraph of type-1 selectors containing other
type-1 selectors to active servers consists of:
* 7372325 menus linked through
* 8843199 live selectors
Of those menus, only:
* 1178121 have valid type-1 selectors to visited
active servers, and each of them
points on average to
* 7.5062 other menus
The remaining type-1 selectors contain no
link to the active servers visited so far (but
they are linked from at least one of them, otherwise
they would not be part of the graph).
You find below a plot (axes in log-scale) of the
number of menus having a given out-degree (i.e.,
number of type-1 selectors to other menus):
1e+06 ++++----+-++--+----+-++--+----+-++--+----+-++--+----++++
O+. + + + + ++
100000 +-+O.O. # menus by out-degree +..O..+-+
++ OOO ++
++ OO ++
10000 +++ OOO O +++
++ OOO O :O ++
1000 +++ OOOOOOOOOO +++
++ OOOOOOO ++
100 +-+ OOO: O +-+
++ OOOOO :: ++
++ OOOOOOOOO ++
10 +++ OOOOOOOOOO +++
++ + + OOOOOOOOOOOOOO OO + ..O ++
1 +-++----+-++--+----+-++-OOOOOOOOOOOOOOOOOO-OOO-+----++-+
1 10 100 1000 10000 100000
outgoing edges
This means that the vast majority of menus in
Goperspace have less than 10 type-1 selectors
to other live menus.
Of those 7372325 active menus:
* 7072600 are referred by 1 other menu
* 150873 are referred by 2 other menus
* 77778 are referred by 3 other menus
* 24063 are referred by 4 other menus
...and only:
* 47011 are referred by 5 or more menus.
As a result, each menu in an active server is
referred on average by:
* 1.1995 other menus.
But 97.9% of all the menus in the Gopherspace
are referred by less than 3 other menus.
You find below a plot (axes in log-scale) of the
number of menus having a given in-degree (i.e.,
number of other menus pointing to them):
1e+07 O+++----+-++--+----+-++--+----+-++--+----+-++--+----++++
++ + + + + ++
1e+06 +++ # menus by in-degree +..O..+++
++: ++
100000 +++O.O. +++
++ O ++
10000 +++ O +++
++ OO O ++
1000 +++ OOO +++
++ OOOOO ++
100 +++ OOOOO +++
++ OOOOOO O ++
10 +++ OOOOOOOOOO +++
++ + +OOOOOOOOOOO OO O + ++
1 +-++----+-++--+----+-++--OOOOOOOOOOOOOOOO+-++O-+----O+-+
1 10 100 1000 10000 100000
incoming edges
__________Strongly-connected_Components__________
A strongly-connected component of a directed
graph is a set of nodes such that any node in
that set can reach all other nodes in the set
through a directed path *and* can be reached by
all the other nodes in that set.
The 10 largest strongly-connected components of
the graph of type-1 selectors among active
servers contain:
@@@@@@
* 247302 nodes
* 133193 nodes
* 6221 nodes
* 6081 nodes
* 3768 nodes
* 3398 nodes
* 3333 nodes
* 2020 nodes
* 1987 nodes
* 1914 nodes
You find below a plot (axes in log-scale) of the
size of each strongly-connected component against
its rank:
1e+06 +++--+-++-+--+-++-+--+-++-+--+-+--+--+-+--+--+++--+--+++
++ + + + + + + ++
100000 O-+ component size by rank +..O..+-+
++ ++
++ ++
10000 ++O.O +++
++ OOOOO ++
1000 +++ OOO +++
++ OOO ++
100 +-+ OOOO +-+
++ OOO ++
++ OO ++
10 +++ OO +++
++ + + OOOOO + + + ++
1 +-+--+-++-+--+-++-+--+-+OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO+
1 10 100 1000 10000 100000 1e+06 1e+07
component rank
___________About_Gopher_Archives__________________
It is worth noting that a large fraction of all
the menus found by `burrow` are inside large file
and news archives/mirrors. Typical examples
include ftp.icm.edu.pl, taz.de, bkandbc.com, and
hngopher.com. Those four gopherholes alone
account for about 90% of all the links traversed
by `burrow`. ftp.icm.edu.pl alone (an archive
containing software, operating systems, packages,
and documentation) accounts for about 79% of all
the links in the Gopherspace.
I personally think that this is totally fine,
and actually expected, since Gopher is first and
foremost a distributed file system. The fact
that the "social" aspect of Gopher, with phlogs
and personal gopherholes, is not that central in
terms of bare size does not mean anything: that
part can still be the most important one to
most of us, the humans who make and shape the
Gopherspace ;-)