如果按照静态的标准来做，可以考虑制作多种可替换部件，此时无论使用何种材料应该都能通过螺丝牢固固定，同时减少活动刮擦而减少对外观的影响。

一些因设计需要而一定要活动的结构，可以尝试使用结构方式控制接触点从而控制磨损的区域，也可以作为一种视觉效果

Yiming's News 2026/06/26

[Read More]

Subsequent Works

For layered mesh, two main problems are somewhat complex to solve:

• An intuitive and easy-to-operate method is needed to enter and exit from editing base faces of layers.
• Due to the recursive nature of current evaluation algorithm, a mechanism is needed for layer boundary to store intermediate point result for each subdivision level.

In terms of display, current main issue is the complex tagging logic has very quickly made the shader hard to read, at some point there's gonna be a rewrite to expand bits of it for it to be readable. Operationally, layer enter/exit interaction may adapt single-hide-all-show logic as an easy starting point, otherwise it is quite hard to determine what parts to show/hide based on connected selections are often nondeterministic in terms of desired region.

Besides, the snapped structure between layers may not be able to be visualized very easily from a topological manner. So far a doable way is to draw snapped vertex positions while transforming (may include deviation lines, and also allows for one time snapping). This seems to require each vertex to store additional snapped location, currently it's unsure whether this part of the memory needs to be dynamically allocated.

How to record point positions for all levels of subdivision for layer boundaries: So far the intermediate IDA implementation might be the most efficient, but if this is used to solve such problem, dynamic length support needs to be added to vertex/edge IDAs, and on top of that, intermediate IDAs only relocates one level of target representation, and that's not necessarily the highest level of subdivision. And intermediate positions needs to be recorded up to the highest level of subdivision. Directly allocate tiny bits to each of the vertices and edges that requires such data is definitely doable but fragmentation can be an issue. Since each one of the additional allocation is of the same size, there might be space for optimizations (like just use memAcquire).

When these two problems are solved, it should be possible to implement layer snapping and inter-surface interpolation. It seems that the same kind of problem of how to record and preserve final point locations for faces still exists, but in this case only positions from highest level of subdivision is required as opposed to from every single level.

研究 Script8 作为嵌入文本上下文中的程序控制语句，可以用作排版

727操作手册

https://digitalcollections.library.miami.edu/digital/collection/asm0341/id/224599/rec/2^↗

看来生成PDF并不困难，嵌入文件等操作也很简单。

从PDF1.2开始支持Unicode，PDF1.3开始就支持图片带ICC配置文件，从PDF1.4开始支持透明对象。

可以自己编写一个小程序用于简单的PDF生成，可以用于制作网站/文档以及相册。

删数据块的时候 NoLongerExists 似乎没有调用过，会导致UI元素解引用时使用删除的数据

感觉可靠性比较差，而且视觉上呈现的效果可能跟宽倒角差不多，不一定好看

如果需要在 ufw 添加前置规则，需要插入而不是直接添加。

sudo ufw insert 1 deny from xx.xx.xx.xx to any

Yiming's News 2025/05/27

[Read More]

Subsequent Works

The main stuff is to achieve a solid G2 evaluation. If accurate varying normal is required, this can become very much computationally heavy, some simplified algorithm is potentially needed.

Then the ring-cut / cut / middle control point slide feature should be implemented / migrated to better verify the correctness of this modified Catmull-Clark surface evaluation algorithm.

Leg

Qooban26d6

Yiming's News 2026/04/25

[Read More]

Subsequent works

Currently the way "Wu-Surface" handles topology at 3/4+ stars is not ideal, but solution already exists, waiting to be implemented.

Soon file r/w and undo support will be implemented as well.

The implementation of T-Junction still has some additional issues. Although it worked logically, but the interpolation method on each side of the T-Junction still needs a more solid definition. Then we can look at doing layered topology.

预测，在接下来的十年内，手动直接控制的机械式功率开关/阀门装置将更加普遍地出现在高可靠性控制系统的设计里。功率器件小型化，高强度塑料以及新的密封技术将允许将带有蓄能装置的微型功率开关整合进合理尺寸的控制面板。