Tom Lane wrote:
> Naz Gassiep <naz@mira.net> writes:
>
>> Joshua D. Drake wrote:
>>
>>> Example discussion with customer:
>>>
>> ...
>> Finally, in the absence of security concerns or performance issues (and
>> I mean the "we can't afford to buy better hardware" type edge of the
>> envelope type issues) there is zero *need* to upgrade.
>>
>
> This line of argument ignores the fact that newer versions often contain
> fixes for data-loss-grade bugs. Now admittedly that is usually an
> argument for updating to x.y.z+1 rather than x.y+1, but I think it
> destroys any reasoning on the basis of "if it ain't broke".
Not when you consider that I did say "in the absence of security
concerns". I consider the possibility that a bug can cause me to lose my
data to be a "security concern". If it's a cosmetic bug or something
that otherwise does not affect a feature I use, then upgrading, as you
say, is very much of a x.y+1 wait than upgrading minor releases
sometimes multiple times a month.
It must be remembered that human error can result in downtime, which can
cost money. Therefore its a foo risk vs bar risk type balance. At least,
that's how I see it.
