On 14959 March 1977, Pirate Praveen wrote:
>>>> 8 node modules (all of them dependencies of gitlab) depend on this
>>>> module (colormin, cssnano, eslint-import-resolver-webpack,
>>>> eslint-plugin-import, postcss-merge-idents, postcss-minify-selectors,
>>>> postcss-reduce-transforms, postcss-zindex). Do you think duplicating
>>>> this code 8 times or maintaining 8 patches are better approach than this?
>> For one line code: yes. You can never change this one line anyway.
> I'd like to ask other ftp masters if they agree with this assessment.
> How do we know for sure we will never have to change this line? Even if
> I agree this need not be changed, is having duplicated work 8 times
> still better compromise?
Duplicate of one line? Yes.
Leaving out the insanity of ways that node goes with its design of
idiotically small modules for things, lets consider what it means in
A one line, no matter if its 20 or 1000 characters long, needs a
package. Upstream with license, readme, maybe tests, in Debian then with
all the stuff in debian/, most files larger than the actual code.
Then you end up with the .deb. Now that and the source get onto mirrors.
And into all packages files, blowing them up. And needing to be
processed in all the various locations all over.
So we have 114 bytes of code and 50 to a hundred times that size of
metadata, on every mirror. And one more stanza in the packages file to
parse for every tool, yet more data and time at totally unrelated
Thats why we ended up long ago to deny small packages. Yes, most of node
is small, so we really dislike it, but if its at least a bit more than a
line, we grudgingly accepted it. (Also, actually most of node is unfit
for packaging, neither is the nodejs system made to be packaged, nor is
Debian made to handle this stuff).
I think reject faq needs to make it more clear on size, currently its
included in "Package split". I'm unsure what a good value is for minimum
size. It can't be one value. But 114bytes of source? I'm with