The narcissism of small code differences

Chris Angelico rosuav at gmail.com
Mon Nov 11 10:53:18 CET 2013


On Mon, Nov 11, 2013 at 8:25 PM, Mark Lawrence <breamoreboy at yahoo.co.uk> wrote:
> On 11/11/2013 06:50, Rick Johnson wrote:
>>    In a nutshell the author attempts to plead for the
>>    "longevity" of "old code bases" simply on the basis of his
>>    assertion that "old code bases" are "less buggy" and
>>    contain more "wisdom" than their new brethren -- both of
>>    which are absurd conclusions!
>>
>
> I recall that the demise of Netscape was due to them trying to completely
> rewrite code of this nature.  The exception that proves the rule?

No, I would say Netscape's experience is a strong supporting example
of the contrary position. I've frequently had need to reimplement
something (most recently as part of porting functionality from RosMud,
written in C++, to Gypsum, written in Pike), and referencing the old
code is the best way to benefit from X years of corner-case
discoveries and bug fixes. Sometimes I'll even copy and paste a
comment from one version to another, because it perfectly explains
some oddity that the code has to handle - even if the code itself is
completely rewritten. Definitely if hunks of code can be kept, they
should be. The old code base *IS* less buggy and contains more wisdom.
Hardly absurd.

ChrisA



More information about the Python-list mailing list