BrianZill
BrianZill
The stack currently only supports a single network interface, which is expected to have a single MAC address and a single IPv4 address. Even if we decide we only care...
This is about code clarity. Currently, the stack has several structures referred to as "options" that are really configuration parameters. This can be confusing, as some protocols (particularly TCP) have...
PR demikernel/inetstack#193 (and runtime's PR demikernel/inetstack#34) introduced dynamic dispatch for PacketBuf (as part of the removal of the runtime generic). This has the usual problem of dynamic dispatch in Rust,...
## Context Currently, the PAL is organized by type of thing being abstracted: constant, function, structure, etc. And/or by platform (there currently is a linux subdirectory). ## Proposed Solution The...
In the current Catnip LibOs, once we've decided to send a packet, we always allocate a new DPDK MBuf to hold the headers. Then depending upon the size of the...
Current Situation =========== In the current API, `demi_wait_any()` closely resembles the "select" API from BSD Sockets, which was designed before multi-threading (or async) was common. It allows you wait on...
We currently don't support Path MTU Discovery, which is a technique for avoiding in-path packet fragmentation. See [RFC 1191](https://www.rfc-editor.org/rfc/rfc1191) (or [RFC 8201](https://www.rfc-editor.org/rfc/rfc8201) for the IPv6 version).
## Context [RFC 6528](https://www.rfc-editor.org/rfc/rfc6528.txt) recommends using a cryptographic hash function, along with a (temporally constant) secret key, in the generation of a TCP connection's initial sequence number (ISN). Our current...
Back in olden times, when dinosaurs roamed the earth and floating-point units were on a separate processor board (e.g., VAX 750), TCP/IP stacks did all of their retransmission timeout calculations...