SDSL Analysis

Benefits of using SDSL data structures including WTs in SeqAn3:

• similar speed, but much smaller indexes
• less work for us

Open Questions regarding SDSL features:

• smaller alphabet support necessary?
• mmapped vectors and data structures?
• bi-directional index?
• construction algorithms? We have fully parallelized in-place SA construction
• how does serialization work? is it platform independent?

Future of SeqAn indexing module:

• hard dependency on SDSL
• q-gram indexing and other non SDSL indexes (WOTD? DFI?)
• new find interface around SDSL indexes, seeding strategies (01*0, kuchererov...)
• parallelization

Benefits to SDSL:

• larger userbase, more visibility (Intel, de.NBI)
• larger platform support
• more automatic and regression testing
• more contributores

Necessary changes to SDSL so we can use it:

• switch to MIT License
• support all seqan compilers + platforms (especially VS, icc...)
• same warning/pedantic level

Political questions:

• formally announce cooperation?
• will design decisions be decided together in the future?
• will some of our developers get repository access?