Why not remove old stuff.

hruodr said:
Why? To save Desktopusers that think in Giga and Terabytes few MB of disk?

Sendmail is called a "big MTA" in than man page, but it contains very few MBs.

The same question: why to delete the venerable standard editor ed?

This proposals come from people that do not have a feeling for Unix culture, but only want a cool operating system.

And I like that FreeBSDBSD is a "Software Distribution" containing a lot of usable old programs. The ones as packages, the
other as the ones expected from a Unix and BSD system, because they were always there.

Crivens, not "Orrrrderrr!!", but O::: de:::!!
Click to expand...

Because the base system is supposed to be useful and efficient for the majority of the users, and how many users do you think use or need a complete mail solution like Sendmail? And those who use/need that, just a fraction of them use Sendmail... There are tons of others mail servers out there and a lot of people use them instead of Sendmail.

Also, it is not like all or the majority of the server installations run a complete mail solution on them - more like just a fraction of them. So, that has not to do with desktop users. Desktop is just an appetizer on FreeBSD, nothing is done to accommodate desktop users if that will penalize the server users, specially the professional ones.

With your mentality I think you will shocked with some things I think will happen in a not so long feature. Old software should just be kept when they are useful and when there is no new but proven superior solution for the given user case in the market, but today there are several mathematically proven superior solutions out there for several user cases, but none of them is in the average Linux market, and so the average programmer don't know about them or don't have proper ways to deploy them.
 
rigoletto@ said:
Old software should just be kept when they are useful and when there is no new but proven superior solution for the given user case in the market
Click to expand...

If that is an argument, we should stop using and developing unix/BSD/Linux and change to something superior, for example plan9.

Many (most?) of us are used to unix and are unwilling to change exactly as Windows users are used to Windows.
If you make the system unrecognizable, then you help others to search for a better OS.
 
hruodr said:
If that is an argument, we should stop using and developing unix/BSD/Linux and change to something superior, for example plan9.

Many (most?) of us are used to unix and are unwilling to change exactly as Windows users are used to Windows.
If you make the system unrecognizable, then you help others to search for a better OS.
Click to expand...

As far I know Plan9 is not really actively developed anymore.

[EDIT]

Also, who is us? The people who actively develop the system, those actually funding the development, the ones who use professionally, regular contributors, desktop users, or many others groups?

You can't accommodate everyone without heavy compromises, this is impossible, and so when you need to choose you go with the ones who contribute more and make the things happens. This is exactly why Linux became what is became, doing all necessary compromises to accommodate everyone, all fashionistas.

Most people think Linux is king. Yes, it own the majority of the toy market but Linux don't even feel the smell of any true critical systems nor any other open-source OS but Muen, those are all running stuff like AIX, VxWorks, custom made OSs and a few others alternatives. Well, we could talk about Linux in mainframes or supercomputers, but that is a completely different Linux, and still the actual software running in mainframes is written in Cobol and so not Linux.
 
I am in the mailing list and it is active. There are also some derivatives. Plan9 people were never very loud.

And even if plan9 is not unix, much modern: it keeps some similarities. Of course ed, lex, yacc, troff, etc are there, although there are "better" substitutes.

One expect from Unixoids similarities. If FreeBSD or something else begin with big changes, it is not more Unixoid, but perhaps a KDE or Gnome machine.
 
Zirias said:
That said, I don't think many early BSD contributors would support your point of view. Why did they start BSD in the first place? They wanted change and improvement in AT&T's Unix.
Click to expand...
Your summary is not completely wrong, but also not right.

Early on, Unix only existed at Bell Labs, and then AT&T licensed it out, mostly to academic institutions. One of those institutions was UC Berkeley, more specifically the computer science department. Once Berkeley became involved in Arpanet-related research (remember, the founding institutions of Arpanet didn't include Berkeley, but instead UCLA, Stanford Research = SRI, UC Santa Barbara, and Utah), they wanted to do research on improving the networking stack of an OS. To do that, they used the AT&T Unix license they have, heavily modified the networking stack, and also added other convenience features (for example "Control Z" to interrupt and then resume a program), and called it the "Berkeley Software Distribution", namely a modified version of AT&T Unix.

What is quite symptomatic: The group that did that was called CSRG, which stands for Computer Systems Research Group. Their goal was not to make AT&T Unix better. Their goal was to do research on computer systems, in particular network-attached ones. For that they needed an OS. And instead of writing one from scratch (which a lot of other people were doing at the time), they simplified their life by starting with the AT&T license they had. The fact that they happened to improve Unix in the process was a by-product, one whose fruits we are still enjoying today.

To the other topic of discussion here, sendmail: When the CSRG folks needed a way to get a sensible and configurable software stack for transmitting mail, it is quite logical that they turned to a family member of a senior CSRG researcher and BSD developer. This is how it happened that Eric Allman programmed sendmail, and it got tied into BSD. While I personally don't enjoy configuring sendmail, and use a different MTA on my machine, we need to recognize and respect that sendmail and BSD have travelled a long road together. And that lots of systems rely on sendmail, and abandoning those systems in an upgrade cycle would be a problematic decision.
 
But it can't be the purpose of an operating system to become a museum.
Proof : You can easely do chmod 000 csh sendmail syslog and I don't know what and find out nothing is really dependent on it.
 
Alain De Vos said:
But it can't be the purpose of an operating system to become a museum.
Proof : You can easely do chmod 000 csh sendmail syslog and I don't know what and find out nothing is really dependent on it.
Click to expand...
If it works for you, by all means, go ahead!
On my system, you would have 50% success: Root would never log in again (fortunately, I have another root account that uses bash, so I could get around that). On many other systems, no mail would enter or leave again. You may not be using your system for mail (*), or you may be using an alternate MTA (I do).
The important thing is: Do not extrapolate from your system to all others. Neither csh nor sendmail are museum pieces, they are in everyday use by many people.

Actually, there is a well-known joke about that. Sendmail was written initially and is still chiefly maintained by Eric Allman, who is openly gay (and a very nice and cool person, as is his better half Kirk). Even today, sendmail handles a large fraction of all mail worldwide (it is getting less, as the largest mail providers are switching to their own custom-built systems, but those often start with the sendmail source code). The joke from Eric is: It is funny that even an ultra-conservative fundamentalist can not send some bigoted hate-mail over the internet, without it being touched by a gay program somewhere along the way.

Footnote *: Even if you don't transport normal everyday user mail, you should have a functioning MTA on the machine, and make sure e-mails to root go to some real human being, because reading the messages from various nightly / weekly / monthly tasks is a very good idea, and will make most systems live longer.
 
Alain De Vos said:
Proof : You can easely do chmod 000 csh sendmail syslog and I don't know what and find out nothing is really dependent on it.
Click to expand...
I'm sorry but your proof is severely flawed. All my users use tcsh. My systems would be completely bricked. It's just too easy to say something isn't good or useful because you don't use it yourself.
There are multiple shells from which to choose, syslog is incredibly important and sendmail is used to deliver mails from running cron jobs. I do not use ksh, zsh nor even bash.

Also, I don't really get what the discussion is really about. If you don't like something, use something else. Vi to me is unusable, some people trust it so much that they'd risk their lives on it.
It's just what it is. One person wants A, another wants B and yet another wants Z. There's a reason all these shells, tools and so much more have been written. They're written by people who
wanted to use them, they're shared by people who believe it might benefit other people. Respect that, but use whatever works for you.

FreeBSD is about options. FreeBSD is and hopefully always will be free. Not free as in without costs, but free as in freedom. Freedom comes with choice, freedom comes with responsibilities.
Whatever is the case, freedom should never be questioned. You do with your freedom as you will, I will do as I will. And we all lived happily ever after! :cool:
 
Offcourse there can be situations where there is dependency on "older" software. But not everywhere or not as much as some claim to be.
I also never said something is bricked or flawed nor that freedom ought to be questioned so this remark falls out of the air.
 
If it bricks my system, I'd call that a dependency. Then again, maybe we're getting lost in semantics now?
Edit: not fair that you kept editing your message, now my reply seems non-sensical
 
Alain De Vos said:
Offcourse there can be situations where there is dependency on "older" software. But not everywhere or not as much as some claim to be.
Click to expand...
Contrary to Microsoft or Apple's marketing, "old" software is much more critical to everything you do on a computer than you think. Around 90% of the code running on Windows 10 was written over 10 years ago.

Also based on the "Lindy Effect", the older a piece of software or technology is, the more likely it is going to be around much longer than a newer bit of software. Possibly a good example of this is the C programming language and GCC compiler. This has outlived many more recent "trendy" compilers and languages. You will also see this in the future with Xorg and the X11 protocol outliving whatever the heck Wayland thinks it is.

Try not to think of this kind of software as "old", think of it more as a standard that has a feature complete implementation that cannot really be improved in any meaningful way simply with a rewrite.

If instead you want a complete modularization of FreeBSD like Linux (with a vague concept of a base) then that simply will not happen. It has proven to be largely a failure contributing to the fairly "wild west" broken landscape seen in Linux distributions.
 
ralphbsz said:
And that lots of systems rely on sendmail, and abandoning those systems in an upgrade cycle would be a problematic decision.
Click to expand...

I started using FreeBSD in 2016 and the willingness to remove sendmail was not new at that time, then DMA got in. I think that was on 10R (I started with 10.3R).

For me that was already a clear indication some decision about sendmail was already more or less to be made. If I depended on that I would have switched to ports already.

The problem of this sort of situation is the fact users need to be actively watching what is going on at development side to catch the intentions early. There is a lack of formal scheduling easily available; however is difficult to have something like that given the lack of others things too, but I know decisions were made, there is work being done at the infrastructure side, I think by someone we both know and trust for that, and I am confident we will get in there, sooner than later. :)
 
When someone claims freebsd is not modularised :
ls /boot/kernel | wc -l
Let's also not forget my home PC must be able to serve a total 2 mailboxes :). A common situation when one prefers more readable config files.
Yet a question comes to my mind. Do freebsd, openbsd and netbsd take a different approach on what is in the base system or are they very alike on that part?
 
PMc said:
The ed is scriptable. It may be used in whatever shellscripts behind the curtain.
And people are still asking for such kind of tools.
Click to expand...

Actually "ed" has chances to be deleted one day, as it is in Linux. Who wants to use "ed" today? In some Linux distributions, there is no "vi" (default installation), but instead likely "vim.tiny", or "nano", or just "vim". I guess that the oldest word processing, like "ed", should be the standard and there by default. "ed" is clean. In addition, the way how "ed" works is very efficient. Keeping old Unix software stuffs has some good.

However, because many users are used to Android and click-and-go softwares or technologies, then, software(s), which haven't an easy-to-use user interface, are likely to be forgotten. It requires to (re)learn and to have time for console based softwares.
 
ralphbsz said:
The fact that they happened to improve Unix in the process was a by-product, one whose fruits we are still enjoying today.
Click to expand...
Well, that's somewhat debatable. Sure, working on software at university is done out of academic interest. But, on the other hand, that academic interest is the search for improvement (through better knowledge).
ralphbsz said:
we need to recognize and respect that sendmail and BSD have travelled a long road together
Click to expand...
So? respect is probably fine here, but concluding that "sendmail can never be removed / replaced" is exactly the kind of thought I wouldn't expect around BSD :) The question should be -- is sendmail still the best piece of software to complete a base system as its MTA?

(edit: or maybe even -- should a base system really still have an MTA or wouldn't it be wiser to have a simple MDA here?)
 
Zirias said:
The question should be -- is sendmail still the best piece of software to complete a base system as its MTA?
Click to expand...

Sendmail is better than its fame. Perhaps much better. With m4 macros it is not anymore difficult to configure.
It is well documented. Gives usable log. It is old and you get answers when you google. I have it running
for relaying with authentication and it was not difficult to configure.

The ammount of disc cannot be an argument: the tar.gz source has about 7 MB.

But I do not know reliable comparitions about security, performance, scalability, etc.

In the time as any MB disk was precious, we had MTA and DNS server in the base system. And now, when
we throw GBs away, we are counting MBs for these software pieces.
 
Alain De Vos said:
When someone claims freebsd is not modularised :
ls /boot/kernel | wc -l
Click to expand...
I was referring to base software, not kernel driver modules.

Alain De Vos said:
Yet a question comes to my mind. Do freebsd, openbsd and netbsd take a different approach on what is in the base system or are they very alike on that part?
Click to expand...

They are semi similar in that they have a concept of base which is pretty stable. The big difference I note is that OpenBSD includes an Xorg fork (though they claim it isn't a fork because they upstream their changes) called Xenocara. This is because they recognise that they can lock it down and secure it best if they don't have to depend on the upstream. They also made its build system monolithic to simplify things. I agree with this; Xorg should be 1 package rather than hundreds. However I possibly would not like it in base if I really didn't need graphical output. You can remove it from base (during installation) but they recommend not to.
 
You must log in or register to reply here.
Back
Top