Developers of Sympa

[Soji Ikeda <address@concealed>]

Marc, at first, would you please compile summary of irc talk?
Please refrain from mixing other discussions and/or your personal thought.  
Thanks.

— Soji

2018/03/09 23:34、Marc Chantreux <address@concealed>のメール:

> Those are the decisions to do about sympatic
> 
> A. Add Path::Tiny to Sympatic
> B. Moo imported by default in sympatic
> C. minimal version of perl
> D. autodie
> 
> decisions A and B will be adopted if there is no more comment.
> 
> A. Add Path::Tiny to Sympatic
> 
> This was a proposal made by eiro on the list but wasn't really debated.
> 
> * eiro thinks we can replace a lot of homebrew code just using
>  methods of Path::Tiny
> * racke thinks we should not provide default unicode policy while using
>  file and we should use Path::Tiny methods instead
> * no one provided negative comments on this proposal.
> 
> i think this should be added as dependency of Sympatic and path() should
> be exported by default.
> 
> B. Moo imported by default in sympatic
> 
> racke proposed to import the moo stuff by default and it makes sense as
> most of the files we have to write are Moo classes. the proposal i made
> (uncommented for the moment) is
> 
>    After               Before
> 
>    use Sympatic -oo;    use Sympatic;
>    use Sympatic;        use Sympatic -noo;
> 
> will be applied unless extra comments.
> 
> D. Unicode policy
> 
> this is still an open topic but some things seems to be clear:
> 
> * there is no good reason to load charnames by default and the memory
>  should be used for more interesting things
> * the default policy of the open pragma which implies use
>  PerlIO::utf8_strict is too intrusive.
> 
> so utf8:all will not be used and we should define our own policy. things
> we have to decide (the way i understand,) are:
> 
> * should @ARGV be encoded by default        ? which layer to use ?
> * should open  use an encoding by default   ? which layer to use ?
> * should stdios use an encoding by default  ? which layer to use ?
> * should the Sympa code be written in utf8  ?
> * should the features fc and unicode_strings be loaded ?
> 
> my answer is: as long as utf8 became the expected format for almost
> everything nowadays, we should expect it as default and using the
> recommendations of the perl pumpkin on it
> (which are described there with a very interesting overview of how
> perl actually handle strings: https://www.youtube.com/watch?v=TmTeXcEixEg).
> 
> so the code and the files should be utf-8 by default:
> 
>    use utf8;
>    use open qw< encoding(utf-8) :std >;
> 
> also @ARGV should be encoded and we should use fc and unicode_strings.
> 
> the last opinion i undestand from Soji is:
> 
> * we should use charnames import
> * we should use :bytes as default open layer
> 
> and i think racke worried about the filehandles but frankly: i really
> think everyone uses utf8 by default nowadays and the other situations
> should be wrap into functions. for example: opening a pipe to sendmail
> by default shouldn't be required from the sympa code as the PEP (perl
> email project) provide modules to do that.
> 
> D. autodie
> 
> the idea was to have a defensive default behavior. it seems a bad idea
> for everyone.
> 
> * no one wants to use a lexical 'no autodie'
> * everyone is pretty sure we will always test every function call that
>  can rise $ERRNO
> * the $ERRNO seems to have a very tight scope so for example: if you
>  try to open a file in an eval block, you loose $ERRNO while leaving
>  the eval block. the consequence is $ERRNO can't be used in a
>  try/catch.
> 
>    try { open $fh,'>','/O B V I O U S - FAIL' }
>    catch {
>        say $ERRNO; # $ERRNO is empty there :(
>    }
> 
>