Crivens said:
The question is if you consider the tradeof as sufficient. If you need reliability and reboot once a year, then this question is easily answered. You do not take that risk. If your typical use involves a lot of rebooting, then the answer might be something else.
As stated before, my systems spend only some seconds between calling
init and offering the login screen, so for my typical usage pattern I have spend more time in this thread than I could save by speeding up the boot sequence. So this is about academica for me.
The boot can be divided into three parts:
How much time is spent in the BIOS can not be changed by the kernel. The time spent in BIOS or Kernel can not be changed by the init system. Speeding up system init can introduce risks for little benefit, speeding up kernel bootup can (at least on my typical use cases) save a lot more time. That is why I use custom kernels, without the SCSI delay and without the drivers I do not need to boot the machine. This may save only little in space in the kernel file, but not probing for the hardware which is not there speeds up the boot process damatically. Maybe this is what you experience.
Thanks for explaining a boot sequence, not really necessary as I already know this, but perhaps useful to other readers. Just because your experience is fine, doesn't mean others are.
Point 1: should be taken into account.
Agreed it should be a factor, but not a reason for not investigating something and trying different options. Using it as a reason for not doing the investigation or trying different ways of improving it is rather poor.
Any change, of any kind, involves risk. If any users are so concerned by that risk then they probably should never upgrade since there will always be a chance that their specific setup will tank.
In any event, anyone truly concerned by stability would never be seen dead running the latest and greatest version of any operating system. The prudent thing to do is to lag behind a couple of releases and wait until its been throughly tested by the wider community.
Point 2: should also matter, but I do not see how that point was made.
I suggest you read back through the thread I don't really want to make this personal.
So your saying that if there is a learning curve (i.e. make some sort of effort) then it should be a reason not to have a change. Well, that could pretty much rule out most changes.
Then again, people concerned about having to learn could always decide not to upgrade and stay with their safe and stable versions that require no extra learning on their part.
Point 3: who made that point?
Sorry, I'm not going to make this personal and name names. Read back through the thread.
Point 4: this point is not accurate I think. Servers typically are rebooted once in a blue moon, thus the savings from speeding up booting are not as vital as taking the risk of bricking the system. Speeding up startup is simply less important.
For
you specifically yes, but not for other server uses - spinning up VM's (ala AWS) would be one I can think of right now. Also, FreeBSD is not just for server uses is it.
In any event since you don't need to reboot that often then you shouldn't mind if its faster and of course since stability is
so important to you then your safest bet would be be several OS's releases behind anyway.
Ahm, I suspect a C&P error here because I do not see how clamav plays into this...
[thread=34581]Try this one, I'd use the wrong bb code[/thread]
I must say that I find the reactions on here quite amazing sometimes - on the one hand there seems to be a huge amount of quite rightly deserved pride in the BSD Community's ability to implement well thought out skillfully designed solutions while on the other hand from some quarters at least an absolute inability to put individual needs aside and as a result outright rejection of investigating how to achieve said skill fully designed solutions to a problem.
According to this:
http://www.freebsd.org/advocacy/myths.html
BSD makes a great server. It also makes a great desktop.
Slow startups on default installs of both PCBSD and FreeBSD would seem to make that statement no longer as valid as it could be if we look at the startup times of others.