From 1cc10ff265ce6a3f7a6a9d23792b02dd03083c13 Mon Sep 17 00:00:00 2001 From: GNU Hurd wiki engine Date: Tue, 11 Mar 2008 21:04:33 +0000 Subject: web commit by hammy: Removed ``translator stacking mechanism''. Already assigned to Carl Fredrik Hammar. --- community/gsoc/project_ideas.mdwn | 41 --------------------------------------- 1 file changed, 41 deletions(-) (limited to 'community/gsoc/project_ideas.mdwn') diff --git a/community/gsoc/project_ideas.mdwn b/community/gsoc/project_ideas.mdwn index 661f8839..c6a721d5 100644 --- a/community/gsoc/project_ideas.mdwn +++ b/community/gsoc/project_ideas.mdwn @@ -8,47 +8,6 @@ Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled [[GNU_Free_Documentation_License|/fdl]]."]]"""]] -* translator stacking mechanism - -One of the major strengths of the Hurd design is that not only it uses a -modular (multiserver) architecture internally, but also encourages a similar -modular architecture at application level. Complex applications can be created -intuitively from simple components by combining them through translator -stacking -- similar to traditional UNIX pipes, but much more suitable for -complex interaction and structured data. - -The downside is that communication between the components with filesystem- or -other RPC interfaces imposes a relatively high overhead: Not only are the -actual IPC operations relatively slow in Mach, but also communication is -generally more complicated because of the constraints of an RPC interface. - -In some cases, like network stacks for example, the overhead might prove a -serious problem; a mechanism that allows cutting down on it by more or less -transparently, taking a shortcut in certain situations to avoid actual RPC, is -highly desirable. - -The work on [[libchannel]] as a special-purpose translator stacking mechanism -as last year's GSoC project yielded some very interesting ideas for a more -generic translator stacking framework. - -(links) - -The task here is to follow up on the previous work, creating a framework based -on these ideas, and implementing or porting some translators based on this -framework to prove it's applicability in practice. It is up to the student to -decide whether it's better to start with the previous libchannel code, turning -it into something more generic; or starting from scratch, using libchannel (and -libstore) only for reference. - -This task is pretty involved. The architecture of the stacking framework has -been discussed before; but the student needs to design the various actual -interfaces for the framework, and implement all of them. - -Last years student might be able to get this done before GSoC starts. If -you're interested in this project send a mail to -[Carl Fredrik Hammar](mailto:hammy.lite@gmail.com) to make sure he has not -already gotten started with it. - * sound support The Hurd presently has no sound support. Fixing this requires two steps: One is -- cgit v1.2.3