Re: [RFC] Yet another option for DPDK options

From: Marc <marcdevel@gmail.com>
To: Thomas Monjalon <thomas.monjalon@6wind.com>
Cc: "Wiles, Keith" <keith.wiles@intel.com>,
	Yuanhan Liu <yuanhan.liu@linux.intel.com>,
	dev@dpdk.org, "Richardson, Bruce" <bruce.richardson@intel.com>,
	"Tan, Jianfeng" <jianfeng.tan@intel.com>,
	Stephen Hemminger <stephen@networkplumber.org>,
	Christian Ehrhardt <christian.ehrhardt@canonical.com>,
	Panu Matilainen <pmatilai@redhat.com>,
	Olivier Matz <olivier.matz@6wind.com>
Subject: Re: [RFC] Yet another option for DPDK options
Date: Thu, 2 Jun 2016 11:19:00 +0200	[thread overview]
Message-ID: <CAExC=0QnX+twj0e5SKviJPmWxbMJRYYRje3krX9znERbOPKFsQ@mail.gmail.com> (raw)
In-Reply-To: <1809080.StTKWmco5F@xps13>

On 1 June 2016 at 20:51, Thomas Monjalon <thomas.monjalon@6wind.com> wrote:

> Hi Keith,
>
> I'll try to bring more context to this discussion below.
>
> 2016-06-01 15:00, Wiles, Keith:
> > Started from the link below, but did not want to highjack the thread.
> > http://dpdk.org/ml/archives/dev/2016-June/040021.html
> >
> > I was thinking about this problem from a user perspective and command
> > line options are very difficult to manage specifically when you have
> > a large number of options as we have in dpdk.
>
> The user uses an application.
> It is up to the application to let users do some configuration.
>
> > I see all of these options as a type of database of information for
> > the DPDK and the application, because the application command line
> > options are also getting very complex as well.
>
> DPDK is a collection of libraries.
> There is no command line options in a library.
> So we should not be talking about such issue. But...
>
> ... configuration of the DPDK libraries must be improved.
> We need some clean API to let the application configure a lot of things
> at runtime (during initialization or after).
> Ideally the API should not use an argc/argv format.
>
> We also have a lot of applications for tests or examples which use a
> common configuration scheme based on command line options.
> It is only for test and demonstration purpose. So it is not so important
> and must not be complex to maintain.
> I also think that we should avoid having to modify a configuration file
> for test applications. I like launching a freshly built testpmd with a
> copy-pasted command line without having to create a temporary
> configuration file.
>
> Instead of wrapping a messy configuration interface, we should proceed
> with this steps (in this order):
> - implement clean configuration API
> - move command line options parsing in a separate library
> - implement an alternative to the options parsing library, as an example
> - remove the options parsing library if the alternative is better
>
>
Fully agree on all that you say. To me:

* +1 on staying away from XML and JSON.

* INI is an option, but if there is the need of hierarchical another option
is libconfig
(
http://www.hyperrealm.com/libconfig/libconfig_manual.html#Configuration-Files)
that to me is more readable and user-friendly than JSON (not to mention
XML).

* As you, Thomas say, and as it has been discussed previously [1]; it would
be good that eal_init was not depending on argv and had a _simple_, and
with reasonable defaults, struct-based init API, and build wrapper
libraries on top of that, one being the command-line and another being a
configuration file (although they would be connected somehow).

Marc

[1] http://dpdk.org/ml/archives/dev/2013-August/000374.html