Part of the really cool things you'd like to have as an app-server author
that knows the APR/APU/HTTPd runtime are subrequests, including the
portions that interact with the filter API.
I'm always writing home-grown app-specific SSI-ish filters via mod_perl to
process application pages as event-driven streams.  It will be interesting
to compare mod_wasm with mod_perl2
in terms of security / sandbox models and weigh them against other
nonfunctional features (memory, latency, CPU, etc).


On Fri, Jan 27, 2023 at 12:28 PM Jesús González <jesu...@vmware.com> wrote:

> Hi Eric,
>
> Thanks for the feedback. Though Wasm is relatively new it is being adopted
> by other HTTP-related projects, like NGINX and Envoy proxy and even ASF
> projects like APISIX. We want to contribute upstream to bring some of these
> new features to the Apache web server, similarly to how we are working to
> contribute our changes to other projects upstream (SQLite, PHP). We plan to
> continue to development and maintenance regardless of future VMware
> involvement. In fact, the reason we started this project was because our
> manager Daniel Lopez was an early ASF member and he looks at this as a way
> to contribute back to the community. We are initially targeting support for
> regular web apps such as WordPress, but we plan for mod_wasm to enable
> safely extending Apache's own functionality, providing access to the module
> API from Wasm code.
>
> On 2023/01/24 20:36:44 Eric Covener wrote:
> > > We are still very interested in contributing this module upstream and
> helping to maintain it. Please, let us know what improvements or changes
> would be needed for it to be considered ready for inclusion.
> >
> > As a pessimistic PMC member not caring about WASM or these languages,
> > I worry that marrying the lifecycle together is not advantageous for
> > either side. Of course I worry about being stuck with the pieces when
> > employer interest wanes or after turnover. It does not seem like it's
> > strictly necessary to be part of the server distribution (there are
> > many examples of successful out-of-tree modules).
> >
> > However the above is no veto.
> >
>

Reply via email to