[dalke]

The README says "PyMOL-RS is a clean-room rewrite" but when I look at ./pymol-mol/src/dss.rs I see things like:

  //! - PyMOL's layer3/Selector.cpp - SelectorAssignSS function
  //! - PyMOL's layer2/ObjectMolecule2.cpp - ObjectMoleculeGetCheckHBond function
  //! - PyMOL's layer1/SettingInfo.h - Default angle thresholds

and "matching PyMOL's cSS* flags from Selector.cpp"

While the Rust code is cleaned up and easier to read, I can see that it preserves similar data flow, uses similar variable names, and of course identical constants.

For example, this is PyMol layer3/Selector.cpp:

          /* look for antiparallel beta sheet ladders (single or double) 
          ...
          */

          if((r + 1)->real && (r + 2)->real) {

            for(b = 0; b < r->n_acc; b++) {     /* iterate through acceptors */
              r2 = (res + r->acc[b]) - 2;       /* go back 2 */
              if(r2->real) {

                for(c = 0; c < r2->n_acc; c++) {

                  if(r2->acc[c] == a + 2) {     /* found a ladder */

                    (r)->flags |= cSSAntiStrandSingleHB;
                    (r + 1)->flags |= cSSAntiStrandSkip;
                    (r + 2)->flags |= cSSAntiStrandSingleHB;

                    (r2)->flags |= cSSAntiStrandSingleHB;
                    (r2 + 1)->flags |= cSSAntiStrandSkip;
                    (r2 + 2)->flags |= cSSAntiStrandSingleHB;

                    /*                  printf("anti ladder %s %s to %s %s\n",
                       r->obj->AtomInfo[I->Table[r->ca].atom].resi,
                       r->obj->AtomInfo[I->Table[(r+2)->ca].atom].resi,
                       r2->obj->AtomInfo[I->Table[r2->ca].atom].resi,
                       r2->obj->AtomInfo[I->Table[(r2+2)->ca].atom].resi); */
                  }
                }
              }
            }

and this is pymol-rs's pymol-mol/src/dss.rs

        // Antiparallel ladder: i accepts j, (j-2) accepts (i+2)
        if a + 2 < n_res && res[a + 1].real && res[a + 2].real {
            for &acc_j in &acc_list {
                if acc_j < 2 || !res[acc_j].real {
                    continue;
                }
                let j_minus_2 = acc_j - 2;
                if !res[j_minus_2].real {
                    continue;
                }
                let acc_jm2_list: Vec<usize> = res[j_minus_2].acc.clone();
                for &acc_k in &acc_jm2_list {
                    if acc_k == a + 2 {
                        res[a].flags |= SsFlags::ANTI_STRAND_SINGLE_HB;
                        res[a + 1].flags |= SsFlags::ANTI_STRAND_SKIP;
                        res[a + 2].flags |= SsFlags::ANTI_STRAND_SINGLE_HB;
                        res[j_minus_2].flags |= SsFlags::ANTI_STRAND_SINGLE_HB;
                        if acc_j >= j_minus_2 + 2 {
                            res[j_minus_2 + 1].flags |= SsFlags::ANTI_STRAND_SKIP;
                        }
                        res[acc_j].flags |= SsFlags::ANTI_STRAND_SINGLE_HB;
                    }
                }
            }
        }

That's close enough that I really think you should include the PyMol license info, before Schrödinger's lawyers notice.

[dalke]

Over the last few days I have learned that using code generation tools are increasingly used to create a "clean room" version of a product, using a definition which is far from its standard use.

See https://tuananh.net/2026/03/05/relicensing-with-ai-assisted-... with discussion at https://news.ycombinator.com/item?id=47257803

I believe your use of "clean room" is another example of misusing the term.

Could you clarify how it was developed? Who had access to the original source code? Were code generation tools used, and if so, how? Was the PyMol source part of the training set for those tools? How did you ensure no copyright violations?

Warren was a friend of mine, and a passionate believer in open source software. He wanted people to be able to modify PyMol for their own purposes, and asked only for a license acknowledgment. Schrodinger, to their great credit, continues to honor Warren by maintaining the Open-Source PyMOL product.

If this project was not developed under true clean room practices, I ask that you continue to honor his work by including the PyMol license in your Rust rewrite.

If it was true clean room development, why does ./crates/pymol-algos/src/align/ce.rs say "This is a faithful port of PyMOL's `ccealignmodule.cpp`.", with comments like "Equivalent to PyMOL's `calcS`" and references to the original code in comments, like: "PyMOL: for (row = 0; row < wSize - 2; row++)"?

[zmactep]

Dear Andrew, first of all — thank you so much for your feedback, both the technical and the legal parts. Your earlier comments about SDF and PDB parsing corner cases are incredibly valuable.

PyMOL has been one of my primary tools for 15 years, and I've always held it in deep respect. This project was born entirely out of a desire to contribute something to molecular visualization in the modern world — something fast, modular, and with qualities I've been missing in existing tools. And as a source of inspiration, I took the best one: PyMOL.

Of course I spent a lot of time reading and studying its code, and I openly took concepts and algorithms from it. I don't hide that — it's why the project carries the name it does, and it's why the README has had an Acknowledgments section since the very first commit: "Inspired by PyMOL, created by Warren Lyford DeLano. This is an independent reimplementation, not affiliated with Schrödinger, Inc."

You are absolutely right about the "clean-room" wording — I used it loosely, meaning "rewritten from scratch in a different language with a different architecture," not in the legal sense. That was misleading, and I've already removed it from the README.

You're also right that DSS or CE was a fairly direct port of PyMOL's algorithm, and it should carry proper attribution. At the same time, many other parts — surface generation, cartoon rendering, the shading pipeline — are done quite differently, and the gap keeps growing. But that doesn't excuse insufficient attribution where code was closely followed.

Going forward, I'm focusing on genuinely new functionality — Rust plugin system, web interface, novel shading models (try set shading_mode, skripkin!) — things the original PyMOL never had. But this is not an attempt to distance the project from DeLano's creation. It's a respectful continuation of his ideas in a completely new product.

Thank you again — your comments are making this project better.