[BlueOnyx:16627] Re: yum update with postfix?

[Larry Smith]

Michael,

  Extremely good, well thought out, and logical explanation.
Sincerely appreciate the time - and thought - that went into
your decision and answer.

-- 
Larry Smith
lesmith at ecsis.net

On Thu December 4 2014 14:52, Michael Stauber wrote:
> Hi Larry,
>
> > Hmmm, seem to recall a discussion not long (*year*) ago about adding
> > postfix as quite a few of us prefer its granularity and filtering
> > capabilities.  Take it we (blueonyx) are now moving back to just
> > sendmail ??
>
> BlueOnyx has always used Sendmail, so we didn't change anything in that
> regards.
>
> But yeah: I explored the option of switching from Sendmail to Postfix
> and had considered Postfix as MTA of choice for 5207R/5208R and onwards.
>
> So I did a bit of exploring and experimenting while I was creating the
> new GUI. The "basic setup" of a Postfix using BlueOnyx 5207R or 5208R
> got hammered out and I stated to write the GUI for it. After progressing
> a bit on that front two things became clear to me:
>
> a.) Porting the existing (old) Sendmail related GUI would be a plain and
> simple 3-4 day job with absolutely no surprises, glitches or problems.
> We'd simply be immediately as good (or better) as we were with the old
> GUI. Simply because the new GUI would use the same Constructors,
> Handlers, Schemas, CODB database definitions and Perl modules as before.
> Just the "visualization" and form processing would be done slightly
> different. The actual trash hauling and heavy duty work would be done by
> the same scripts that we've used for years with great results.
>
> b.) Creating a new Postfix aware base-email module for the new BlueOnyx
> GUI meant starting on a fresh sheet of paper. A basic set of options
> that would need to be configurable needed to be determined. Possible
> values for these options needed to be established ("bool", "integer",
> "word", certain pre-dertermined strings, etc.), the CODB database
> structure would need to be created via new Schema files. Handlers and
> Constructors for that needed to be created. Perl modules to parse and
> edit the Postfix config needed to be created from scratch. So from top
> to bottom the entire "base-email" module would need to be redone. Not
> just the GUI for it, but everything else, too. I had made a good
> headstart there, but just getting the very basic feature set in there
> would mean four to five weeks of coding. And it would still come short
> to the functionality that we already had for Sendmail. Plus I'd need
> two, three or four weeks of testing and bugfixing. And then it might
> just work for me and you (the users) would find things that probably
> wouldn't work for you and that this or that urgently needed a fixing or
> you'd not be receiving/sending any emails.
>
> With that in mind I stepped back, grabbed the shovel, the shotgun and
> then Mr. Postfix and I took a stroll behind the garden shack, where he
> met an untimely end in a shallow and unmarked grave beneath the rose
> bushes. \o/
>
> The thing here was like this: I knew that upon release BlueOnyx 520XR
> would have issues. A project that complex and that big with so much new
> code simply couldn't be expected to run flawlessly from the get go. The
> testing before the release had pretty much eliminated the really massive
> show-stopping bugs, but there were still enough glitches to keep me
> *really* busy for a while.
>
> Had we started with an untried and unproven Postfix implementation
> ("untried" as far as BlueOnyx is concerned), we would have risked to
> ship BlueOnyx 520XR with a potentially faulty MTA. Which is a direct and
> massive show-stopper and would have ruined the usage experience
> massively. You only get *one* attempt at making a good initial "first
> impression". If you blow that, then you're fighting an uphill battle
> from then on.
>
> Nobody really likes Sendmail - that's clear. I don't like it one bit
> either. But there are also BlueOnyx users out there who have never used
> Postfix before, but who (more or less) know how to hit Sendmail on the
> head with a hammer to get it going again. With a potentially buggy
> Postfix implementation these users would have felt kinda helpless or
> even alienated.
>
> So bundling the new GUI with Postfix would for sure have slowed down the
> release of a working 520XR GUI and (on top of it) would have lead to a
> slower adaption by the user base due to more (and more critical) initial
> bugs, which would have resonated with the users even a long time after
> these bugs would (eventually) have been fixed.
>
> Like said: Ruined first impression = uphill battle from then on onwards.
>
> There you have my reasoning for keeping Sendmail. At least for now.
>
> FWIW: Just last night the installation count of BlueOnyx 5207R and 5208R
> (according to YUM updates) went over 500 for the first time. So in the
> 100 days since the release date we've had 500 adopters. That's like five
> a day.
>
> To put that into perspective: In the same timeframe we've had 154 new
> 5106R installs, 63 new 5107R and 523 new 5108R. So just 5108R did have
> more new installs in the last 100 days than 5207R and 5208R combined.
> Which means despite the good first impression people (at least for now)
> are in slight favor of opting for 5108R over 5208R for new installs. My
> hope is that this changes soon, as I no longer see a good reason to use
> the old GUI instead of the new.
>
> Right now I'm working on BlueOnyx 5209R. Again we have a chance there to
> implement Postfix from the get go as only "supported" MTA. This time I'm
> not even trying - for the same reasons. 5209R will have enough issues
> already from the onset. Thanks to RedHat switching to Systemd. We also
> already have a new unwanted step-child aboard in the form of MariaDB
> (instead of MySQL). Plus there are a lot of other fundamental platform
> changes, which (at the current early state of development) throw shovels
> of sand into the gearbox at every turn and bend of the road.
>
> I'd like to keep things as smooth as humanly possible. Hence I'm yet
> again delaying the move to Postfix, as it would rock the boat too much
> for no good reason at all. Yes, Postfix does all the things in less
> arcane manners and (once you know it) with recognizably less effort. But
> at the end of the day it's just the MTA and both Sendmail and Postfix
> get the job done.
>
> So for now let's just stick with "the devil we know". And with that I
> don't mean the old GUI. :o)