Sunday, January 9, 2011

LCA 2011 keynotes: Allman to focus on sendmail

Allman is the father of sendmail, the first mail transport agent for UNIX systems. He will be in Australia later this month as one of the keynote speakers at the forthcoming Australian national Linux conference in Brisbane.

Allman developed the precursor of sendmail, delivermail, in 1981 as an extension to the AT&T Unix code which was available at the University of California in Berkeley. He was studying for a computer science degree.

Sendmail was designed to deliver email over what was then a relatively small ARPANET which had several different smaller networks. Most of these networks had differing headers for email.

Allman's MTA soon became an important part of the Berkeley Software Distribution. It is still widely used on UNIX systems despite being difficult to configure. Alternatives like postfix (written by Wietse Venema), exim (written by Philip Hazel) and qmail (authored by Daniel Bernstein) have gained ground as they are much easier to configure.

But veterans still swear by sendmail. As far as GNU/Linux goes, Slackware, one of the older distributions still uses sendmail as its default MTA though Debian has moved to exim and Red Hat to postfix.

Though Allman, who now works at Sendmail Inc, a company he co-founded in 1998, has also authored software such as syslog (a standard for logging program messages) and several other programs, he is best known for sendmail given its degree of use and the fact that it was the first MTA.

Thus it is not surprising that in his keynote, he will focus on the software that has become synonymous with his name.

"Briefly, my talk is going to explore the architecture of the sendmail MTA from a historical/introspective perspective," Allman told iTWire in an interview. "Like so many other tools, sendmail was originally written as a quick hack to solve an immediate problem; unlike most other tools, it is still around over 30 years later and continues to be one of the major MTAs on the internet."

He said he would first provide an overview of the historic situation. "What were machines like? What already existed in the email world? What was happening that triggered the problem?"

Then would come an examination of design principles and the early days of the evolution of sendmail. "Why did I do it the way I did? How did sendmail change as the world changed?"

Allman then plans to take a somewhat deeper dive into "individual design decisions (as distinct from design principles), including some analysis of whether they were good decisions, bad decisions, or decisions that should have been changed over time."

And to finish, there will be "an overview of what I would do the same, what I would do differently, and what we can learn."

"My hope is that people may be able to take away some knowledge they can apply when architecting a new system," Allman said. "I'm a pragmatist, and a lot of what you read in 'the literature' is disturbingly bogus for actual use in the trenches."

He said that to be honest, there wasn't very much specifically about open source in his keynote.

"The principles I used with sendmail are identical to those I would have used with commercial software - I'm of the school of thought that all software should be written as though it was going to be open sourced, even if it obviously will not, because I think programmers do a better job if they think that others will be evaluating their source code. But that's about as far as it goes."

The LCA 2011 will be held at the Queensland University of Technology from Januarty 24 to January 29.

No comments:

Post a Comment