Pine First-Use Statistics

Beginning with Pine version 3.90 (August 1994), Pine has offered new users a (usually one time) opportunity to send us an email. Initially, the offer was to send for a document called "The Secrets of Pine", which was a brief overview of using Pine and some release-specific information. At some point the document was renamed to "Getting the most out of Pine".

Our not very secret agenda behind these offers was to help secure continued funding for Pine development by demonstrating widespread interest in Pine. We measured this interest by counting the requests we received from new Pine users.

For reasons we were never really able to discover (despite numerous attempts) between .001% and .01% of people who received a Pine document complained that they'd never heard of Pine, never asked for a document, and occasionally threatened legal action for receiving "spam". For this reason, beginning with the Pine 4.0 release, Pine no longer offered anything more than an opportunity to be counted as an anonymous Pine user (via a message we don't answer).

We have no way of knowing for sure what fraction of new Pine users worldwide send us mail when prompted, but we did measure the fraction on our campus timesharing systems and it was about 50%. We feel this may be somewhat representative worldwide, especially if it indicates a complete disregard for the question and a random yes/no selection!

There are some anomalies in our data:

  1. Pine prompts "new users" based on the existence and contents of their ".pinerc" file. If it is unreadable or unwritable or gets deleted, users may get re-prompted. Because it is computationally expensive to test for uniqueness of Pine users, we have not routinely done so and only occasionally sample the replication rate.

  2. Due to a programming error in Pine 3.92 (released March 1996), all Pine 3.92 users (even those who had used an earlier version of Pine) were re-prompted at first use. Checking the replication rate through April 1998, we found that for just Pine 3.92 and above it was 11% and overall it was 16%.

  3. A large number of Pine users are students at universities. As they graduate and their account names get re-used by new students, some apparent duplication is occurring which is not real duplication. We have not attempted to quantify this.

  4. Presumably because of some mailer misconfiguration, an occasional site sends a surprising number of identical requests (37,644 in one episode) and we've now accumulated over 200,000 from "root@localhost.localdomain".

  5. In December 1999, we were surprised to discover the replication rate for the year had jumped to 29% (double that in 1998). We now believe this is mostly caused by programming errors in Pine 4.00, 4.10, and 4.20 similar to that which caused the reprompting in Pine 3.92; although along the way we also discovered that a remarkable number of users seem willing to just press <enter> a few extra times when starting Pine to live with unreadable or unwritable .pinerc files (and the reprompting at every invocation this causes). The 1999 replicates include:
    >1000from each of a dozen very patient Pine users;
    >100 from about 900 users;
    >10 from about 9000 users; and
    >1 from about a quarter million users.
    We are considering changes to Pine to reduce this.

With all this in mind, we currently estimate the number of unique email addresses we've seen to be roughly 33% fewer than the raw number of messages we've received. (But except as stated otherwise, our graphs below are based strictly upon the raw numbers we've received). Also remember that some fraction of users decline to send us email when prompted and some sites choose to disable the option entirely.


running total graph

daily count graph


90% of the first-use email we received came from these domains:

TLD graph


The body of each first-use email also tells us whether that instance of Pine is configured for IMAP and/or NNTP use and whether the user was NEW to Pine (had no .pinerc yet). This graph shows how that has changed over the years:

feature graph


Other things we've learned from receiving these emails:

  1. You can do a remarkable amount of work on a small unix computer if you want to. The autoresponders are shell and perl scripts and the first 20.7 million of these Pine emails were received and handled by a single decstation model 25 running Ultrix. It was about a 25 MIP processor with a physical memory limit of 40 megabytes (about a 486-class machine). Mid-February 2001, the decstation developed problems and was replaced with a (faster/better/cheaper) linux system running the same scripts (with minimal changes).

  2. When one site (or country) has very poor connectivity (99% ping packet loss) they can still eventually get short requests through to us but we are at a disadvantage when we have to send a large reply back. Before we began consolidating requests by site so we could reply to multiple new-user messages at a site with a single message to multiple recipients, we would often find our address space entirely full of sendmails trying to deliver Pine documents to a single poorly connected site (or country). We originally modified our sendmail so that queue runs would learn about unresponsive sites from each other and skip them for a while. Now we simply rely on modern sendmail's host status directory and MinQueueAge features.

  3. When summer ends and a new class of students starts using Pine at schools and universities worldwide, we'd better be prepared to handle the mail. When we first started doing this, we had no idea what that demand would be and every fall we still watch closely. (Note the bump each fall as a new freshman class of students first uses Pine.)

Here is a graph showing the evolution and acceptance of Pine versions. See also the official Pine Release Chronology.

version graph


Finally, below are 6 graphs (in 2 columns of 3 graphs) showing Pine use by host platform as deduced from the three letter code in the Message-ID of the email we receive. In each of the 2 columns:

LEFT RIGHT
SOL Solaris BSI BSDI
GSO gcc Solaris ULT Ultrix
SUN Sunos SV4 SVR4
A32 AIX 3.2 BSD BSD
OSF Digital UnixNXT NeXT
LNX Linux BSF Free BSD
HPP HP PCW Windows 16bit
SGI SGI WNT Windows 32bit
A41 AIX 4.1 PCN Novell

These graphs show that BY FAR, the most common platform is Linux (see lower left graph) but because of the large number of users timesharing Sun, Aix and Digital Unix boxes (see middle left graph), most of the users are still on those (see top left graph). You can also see which platforms are used most by schools because they have the telltale bump at the end of summer each year.

top-left-graph top-right-graph
middle-left-graph middle-right-graph
lower-left-graph lower-right-graph