🚀 go-pugleaf

A mail-based archive server is an impolite and destructive way to
distribute large amounts of information, and should be discouraged in
such applications.

When a mail-based archive server (MBAS) sends a requested chunk of
stuff (CoS) to the requestor, it has no way to know what transport
mechanisms will be used along the route.  If the transport mechanisms
are all on zero-incremental-cost-per-use networks, all is well.  Very
often, at least some of the trip will be along edges in a graph of an
intermittent-connection, store-and-forward network, usually involving
a nonzero incremental cost for transporting each CoS.  Most such edges
are maintained, and paid for, by the nodes on either end.  They
voluntarily allow other traffic to pass over their link, usually with
the understanding that the volume will be small enough to be of
negligible cost.

However, if someone starts shipping megabytes of archives across those
links, the distribution cost is borne by someone other than the
requestor.  This is impolite.  Eventually, such hospitality abuse can
cause the eventual removal of that connection from general "public
service", and the two nodes will return to maintaining a private edge
for their own use.  Traffic will then shift to other edges, increasing
their burden and encouraging them to retreat similarly.  In the
extreme case, the practice of general use and hospitable availability
of connections will disappear.  Everyone would need to maintain
distinct links with every other site with which they wish to exchange
traffic.  Free, open, cooperative communication as we know it would
wither, a nostalgic remnant of a bygone era.

But take heart, there's another way!  Instead, sites can pay for the
connectivity they use for shipping large CoSs.  If they're connected
to a zero-incremental-cost-per-use network then they pay for the
connectivity every month in their leased line bills.  This is the case
on the IP Internet, the BITnet, SPAN, and others.

If a site is not so connected, then they can pay for the transient
connectivity when they need it.  For example, the Computer Science
department at the Ohio State University has for several years
maintained just such an archive.  Similarly, UUNET Communications
Services offers dialup UUCP file transfer service even to sites
without pre-established UUNET accounts, through their "900 number"
facilities.  The connectivity appears in the requestor's next
telephone bill because, unlike OSU CIS, UUNET must recover the costs
of providing the service.  Such archives' contents are freely
available to anyone who wants them and is willing to pay the cost to
acquire them.  No provision is made to assist requestors in
freeloading on other sites' generous hospitality, other than in
exchanging relatively small mail messages containing instructions and
assistance in using the archive.

Any number of schemes are possible, usually based around a reliable
file transfer protocol.  Pay-per-use archives run today using Kermit,
UUCP and Fido request protocols, and probably innumerable others.
Note that in saying "pay-per-use" I mean paying for connectivity
(usually the phone company), not paying for the privilege of access
nor for the archive contents themselves.

The work involved in maintaining such access to an archive is
negligible, compared to the work of maintaining the archive itself.
In fact, compared to the work of untangling bounced mail replies from
a MBAS, it should be quite attractive to the archivist.  Once set up,
the access mechanism pretty much runs itself.

But MBASs are so popular, and so useful to so many people, I often
feel like a voice crying out in the wilderness.  Please, at least
provide and encourage other, more polite distribution means.  I
wouldn't want such a good thing as people's desire to share to be a
contributing cause to the dismantling of open networks!