Re: First pass at filesystem and OS steps note

Alex Milowski <alex@milowski.com> writes:
> 1. pf:copy should allow copying directories.  If it does, it would need to have an option for
> recursive copying.

Uhm. Ok.

> 2. More complicated operations for copying and deleting use regex matching.  Are we going to
> consider this?  While a slippery slope, copying/deleting by extension is a common operation.

Following on the model of p:directory-list, I guess we should allow
include-filter and exclude-filter.

> 3. pf:mkdir needs to allow for creation of a whole directory path ancestry if possible (i.e.
> like mkdir -p)

Fair enough.

> 4. Are we really going to create namespaces for every additional set of steps?  We didn't
> create a new namespace for p:template et. al.  What is the deciding factor for a new namespace?

In our exhaulted position as the XProc Working Group, we can add steps
to the p: namespace, but I'm not inclined to add everything to that
namespace.

The file and OS steps seem like a nice, closed separate library. You only
need them when you need them and only in the pipelines that need them.
So I'm inclined to put them in a separate namespace.

> I would really, really like to avoid a careless proliferation of namespaces for steps.  We
> should think hard about how we want to name steps. 

Indeed. The tension being that I don't want a careless proliferation of
"useful in the 1% case" steps in the p: namespace.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh
Lead Engineer
MarkLogic Corporation
Phone: +1 512 761 6676
www.marklogic.com