彙整

bookmark_border虛驚一場……

Posted on by Isatis Crucifer

結論寫在前面：2026 年 9 月 11 日凌晨往前算大約三到四個月左右，這個 blog 可能會不定時封鎖某些浮動 IP 的訪客，然後可能又會忽然恢復。原因：我的防火牆設定寫錯了……

不小心被封的人應該會看到，我這個 blog 有安裝 Wordfence 這個外掛來幫我做一些基本的攻擊防護；用的是免費版，所以一些功能相對陽春，但防那些想要亂試的登入者應該是綽綽有餘了：我這個 blog 只有一個發文帳號，帳號名也不是顯示在上面作者欄的那個名字，而且我還開了 Wordfence 提供的 2FA 驗證，就算猜中帳號也不會有 2FA 驗證碼可以打。

由於 Wordfence 對於一些攻擊會將其 IP 封鎖一段時間，我時不時會上來巡一下看最近有沒有擋到誰，看到有點過份的我會整理一下之後永封。就是這個整理動作出了問題：前一次整理 (大約是五、六月左右的事) 有把一些 IP 相近的攻擊者寫成 CIDR mask 的範圍 (畢竟這看起來要嘛是 botnet 感染了一票附近的電腦，要嘛是攻擊者本身是浮動 IP，那擋掉一小塊應該沒差)，但我記錯斜線後面的數字是數哪邊了：如果我要封 10.10.10.0 到 10.10.10.255 這個範圍的話，我應該要寫成 10.10.10.0/24，表示前面這個 IP 從左邊數過來 24 個 bit 要對，但是我寫成 10.10.10.0/8，以為是指定從右邊數過來 8 個 bit 可以任意；後一種的封鎖實際上封掉了 10.0.0.0 到 10.255.255.255 這一大塊區域，也就是誤封到了許多沒有要封的人。

然後就是在今天 (9 月 11 日) 我要上來看個東西時突然跳了封鎖頁面出來，嚇了一跳的我趕快進後台看有沒有什麼事發生 (這個 blog 是放在 google 雲上面，我可以直接登入機器看狀況)，但在資料庫裡翻了半天也沒有什麼特別的發現；回頭重新整理時可能是碰到正好跳到沒封的 IP 了所以讓我登了進來 (我還怕說是不是有什麼別的地方把我這個網址隨便導向，就藉著我在後台機器上弄了個簡單的小測試證明了這個網域沒被導走)，然後在看防火牆永封設定的時候才突然驚覺：欸這個 CIDR mask 好像怪怪的，仔細看了看才發現我寫錯邊……而且果不其然，有在這個時間附近觸發的規則都是原本該寫 /24 但被我寫成 /8 導致誤封一大票的項目。嘛，本來這個整理只是我不想在 log 裡看到一大票同樣的東西而已，所以就把那一波整理的名單裡 mask 寫錯邊的項目通通刪掉了。

不知道是幸運還是不幸，今年因為各種原因沒在這裡發文章，在這之前唯一發文日期是今年的是六月底左右時我把我的 GNOSIA 心得文翻譯成英文發在這裡 (那篇從動畫在播時就開始翻，林林總總大概有半年左右的零碎時間吧)，所以其他文章的觸及可能沒有太多影響，但那篇英文心得文可能今天開始才能開始累積觸及……我後面其實還積了兩篇心得文要發，但目前都才寫了個三成左右 (因為各種原因……) 所以這個問題一直到今天撞到了才發現。唉～

喔對了，因為那篇翻譯的心得文的關係，這篇會有英文版放在下面。

[English Version] False Alarm…

Let me get straight to the point: In a period of time about 3 to 4 months until September 10th, 2026, this blog may randomly block visitors that are otherwise legit. The reason: I made a mistake in setting up the firewall block rule.

As those who got blocked may observe, this blog uses Wordfence plugin to do some basic attack mitigation. I’m using the free version, so there are some functions lacking, but for protecting random attacks it should suffice: there is only one post account in this WordPress installation, which is not the name shown above; and I also uses the 2FA TOTP authentication provided by Wordfence, so even if someone accidentally “guessed” the account they still don’t have the 2FA code.

Since Wordfence automatically block some attack sources for some amount of time, I would occasionally login to review the block list, and made some blocks permanent if the attack crossed some line. This is where things went wrong: the last time I review the list around May or June, there are some blocks that target some similar IPs (which, I think, were either botnets from neighboring computers, or a “floating” IP that get reallocated however often), so I collected them, rewrote them into CIDR mask form, and made it permanent. But I misremembered the meaning of the number after the slash: if I want to block 10.10.10.0 to 10.10.10.255, I should write 10.10.10.0/24 indicating there are 24 bits from the left should match, but I wrote 10.10.10.0/8 thinking the 8 bits from the right could be anything. This latter rule actually blocked 10.0.0.0 through 10.255.255.255, meaning I blocked way more IPs then I intended.

Then today (September 10th) I want to login to do something, but faced with the block screen of Wordfence. Since this blog is hosted on Google Cloud, I can access the machine from there, so I logged in, rummaging through the database, but found nothing that may be the reason. When I went back to the browser, it randomly unblocked me so I logged in WordPress from the browser (after confirming that no one is redirecting the domain name to somewhere else, with a little check that I can do when I already logged into the back end machine). I then looked at the firewall block rules, and finally realized that my CIDR mask was written wrong. Adding support to this theory, the rules having last trigger time around this time are those masks that should be /24 but I wrote /8, which means the reason I got blocked is probably somehow my “floating” IP hit those too large of a range. Since the reason I rewrote the rule is simply that I don’t want those similar IPs show up, and now I cannot even know whether these bad actors are still there, I simply removed those wrong IP ranges.

Fortunately (or maybe unfortunately), because of various reasons, there is only one post in 2026 before this post: my English translation of GNOSIA review previously written in Traditional Chinese, which I’ve been worked on when the anime is still airing, working in my free time in the span of about half of a year. I think the reachability of my old posts are not that affected, but this English translation probably suffered a lot; it probably only really “shown to the public” starting today. And I have two reviews queued up that I have only written about 30% of them (which is part of the reason I found this out today). Welp.

bookmark_borderMemories of Japanese Jinro Game – GNOSIA

Posted on by Isatis Crucifer

Foreword: This article is the English translation of my original essay, written in Traditional Chinese, back in April 2025, roughly one year ago, when all we have about the anime adaptation is a promotion video. I chose to translate the essay into English now since the anime adoption is received quite well (from both people played the game and not played the game), and the Traditional Chinese article have unprecedented access traffic for such a small blog. This will not be a one-to-one translation of the Traditional Chinese version, mostly to streamline the essay and put focus differently for English-speaking audience; I’ll also incorporate some modifications reflecting the change of things within this year and the anime, but bear in mind the timeframe of the original essay is about a year ago, when we just heard of the anime adoption and only have one promotion video. This is a REALLY long essay (my word count plugin says this has almost 13K words) so enjoy!

Also sorry for the site still in Traditional Chinese; I only occationally post English stuff here.

Spoiler note: This essay contains discussion of the whole game, including the True Ending; the anime adoption goes in a rather different direction before converge into the True Ending, so some discussion will be different. I would suggest the reader either finishes the game or finishes the anime before reading this essay to avoid major spoilers.

Last Chinese New Year1 going back home, I prepared the game GNOSIA in my Nintendo Switch as the filler game between family activities. This title is actually sitting in my Steam wishlist for a long time waiting for it to come to PC, but after some consideration I decided to buy the Switch version:

  1. The genre of this game is known as Mafia or Werewolf, which I am quite familiar with. I was introduced to a genre of anime called “loop fiction” from Higurashi: When They Cry around 2007, then later from the When They Cry community (especially the Umineko fanbase) I was introduced to the Werewolf game. The variant I played the most is the Japanese web version written in PHP, called “汝は人狼なりや？” (Nanji wa jinrou narya?, literally translated as Are you werewolf?, henceforth shortened as Jinro), which I was introduced by this community following Niconico Douga videos featuring the Umineko characters “playing” this game. I’ve been in one of the Traditional Chinese Jinro community ever since, even after some decline when many of the members were too busy IRL and only holding the community together with the in-game chatroom.
  2. At the end of 2024 an anime adoption of this game is announced, and later revealed it will be on air in October 2025. Because of the familiarity with this genre, I decided to play the game first before anime airs.
  3. The reason to choose Switch version over PC version is twofold: I recently bought Switch (yes, just before Switch 2 announced), and as some basic info I gathered indicated, this game should be easily picked up and put away; there’s no need to stick before my PC to play. Although I may buy this game again on Steam for the achievements.

Here’s the overall gameplay of GNOSIA: on the first sixteen or so loops, the game gradually introduces to the player all the roles of the game and all 14 characters; in later loops the player will need collect background information of these characters, to find out why they (and Setsu) is trapped in this time loop.

During the research of this essay, I found an interesting record of this game: In Dengeki Indie Game Awards 2022, held by the Japanese column Dengeki Indie from the website Dengeki Online, which is voted by (I think) the readers of the column, GNOSIA had a grand slam of all the first places, even acquired the first four places in the Character Department! If you think the other titles in the overall department are already really good (OMORI in second, DDLC+ in third, ENDER LILIES in fourth), you should definitely try this game.

Continue reading “Memories of Japanese Jinro Game – GNOSIA”

bookmark_borderAdvent of Code 2025 / Everybody Codes 2025

Posted on by Isatis Crucifer

第四年在這裡貼 Advent of Code 心得了。今年開始 Eric 決定要減輕一點他自己的負擔，所以題目數量減少到 12 天，不過仍然還是在 12/1 開場，每天開題；也就是今年只會到 12/12 就結束了。

目前看起來這個年末的程式設計謎題傳統有其他人繼承了下來，像是去年就開始的 Everybody Codes 它是自十一月的第一個星期一開始，連續四週的週一至週五各開一題——開啟時間似乎是 UTC+1 的 24:00，換算台灣時間是隔天的早上七點；麻煩的是這個時間剛好卡到我上班只好在到公司時偷一點時間出來寫。他這邊的故事風格比較中世紀一點，並且以鴨子做為各種吉祥物和故事元素，算是很不一樣的風格。

Continue reading “Advent of Code 2025 / Everybody Codes 2025”

bookmark_border人與世界的關係——《無法成眠的伊達鍵 – From AI：夢境檔案》心得

Posted on by Isatis Crucifer

《AI：夢境檔案》系列新作，不過打越在忙他的《百日戰記》所以這個外傳不是他寫的，只有做為監修指導而已，寫的是這作的監督山田和也；不過他在看完劇本及試玩後給出的評論是「僕がいなくても『AI』の世界が描けてしまっている。」1這樣的高評價。我玩起來的感覺也差不多，故事的氣氛 (及垃圾話程度) 真的跟打越寫的正作非常像，而且這作還選了個很有意思的題材在 AI 的世界觀底下探討。

本作的特色就是新加入的「脫出部份」，故事當中會有和《極限脫出》類似的脫出環節 (《極限脫出》也是在 Spike Chunsoft 這邊做的，很多製作人員也有參與本作製作)；故事上是咱們的偶像 A-set 不知為何被抓走關在一個看起來像是 UFO 的脫出房間裡，要從裡面逃脫出來。當然《AI：夢境檔案》本家的夢境環節也沒有缺席；雖然這作沒有故事分歧 (只有搞笑的額外壞(?)結局)，不過以作為銜接前面兩作中間空檔的故事來說倒是頗為恰當。

值得一提的是，這一作中，《AI1》的主題「AI／I／愛／eye」2以一個很特別的型式也做為主題出場，但探討的面相和《AI1》的方向不太一樣，或者該說有所互補；不過可能受限於長度和題材本身吧，拿出來的東西只能簡單擦個邊，不過做為開啟討論的目標的話我想足夠了。(不然也就不會有這篇文章了)

前篇《AI2》的心得因為有部份篇幅分給了和打越鋼太郎這個劇作家和他的作品，不只篇幅超長還作了兩重防雷，不過這篇只會專注在這作及它和前面兩作之間的關係，所以只有一次防雷，還沒玩過的就請先離開了。

Continue reading “人與世界的關係——《無法成眠的伊達鍵 – From AI：夢境檔案》心得”

bookmark_border日式人狼遊戲的回憶：《GNOSIA》玩後感及劇情探討

Posted on by Isatis Crucifer

2026/6/27 edit: This essay has an English version! Let’s see if this essay can reach more audiences.

今年過年我帶 Switch 回去時準備了這個遊戲打算做為零碎時間填充用。這款遊戲其實很早就躺在我的 Steam 願望清單等上 PC 了，但後來多方考量還是決定買 Switch 版。主要理由有幾個：

  1. 這款遊戲的主要內容是人狼遊戲，不過不是近幾年流行的那種少人數的狼人殺較偏重對話的設定，而是約莫在十幾年前比較流行的多人數較偏重邏輯的日式人狼。之前我在《高手過招》系列文裡提到過，我從《暮蟬鳴泣之時》開始被介紹進這種輪迴系作品，然後就從 PTT 的 HOT_GAME 版被拉進日式人狼圈子裡，主要是在玩《汝等是人是狼？》(《汝は人狼なりや？》) 這個版本 (以下簡寫為《汝等》)，那時在幾個國內比較大的伺服器裡都有紀錄在 (用的是另一個名字就是了)；最近因為大家幾乎不怎麼有時間能湊得出遊戲需要的 16/22 人來進行遊戲，所以雖然社群還在但其實已經轉化成一般的聊天群了。
  2. 2024 年底放出了 2025 要出動畫的消息（之後確定是 2025 秋番），想說這個題材 (輪迴人狼) 因為上一條的關係算是我很熟的題目，在動畫播出前我自己也來熟悉一下角色吧，順便為四月底小高跟打越的新作準備，所以就入手了。
  3. 會選擇 Switch 版而不是屯很久的 Steam 版則是：一來我前陣子終於買 Switch 了，二來既然這是單人加速版的人狼遊戲，快的話一場幾分鐘就能夠跑完。這樣一來與其在電腦上玩不如在 Switch 上玩，隨時可以拿起放下。不過看在 Steam 版還有成就上可能之後找個時間補上？

遊戲的前十幾輪是遊戲及職業介紹，以循序漸進的方式一點一點增加複雜度，讓玩家熟悉遊戲的進程及一些基本邏輯；之後就是在不斷輪迴的遊戲當中搜集所有登場 14 個人的背景故事，尋找為何玩家和 Setsu 會被留在時間迴圈裡，以及如何逃出的辦法。

在找資料時找到了一個關於這遊戲的有趣紀錄：《電擊Online》的專欄《電擊Indie》在 2022 年舉辦的《電擊 Indie 大賞 2022》中，《GNOSIA》這部作品包辦了作品部門首獎、角色部門前四名 (依序是 Setsu、Raqio、沙明、SQ)，以及劇情部門、音樂部門、獨特性部門三個分部門的首獎！如果你覺得當年跟在其後的作品 (第二名《OMORI》、第三名《DDLC+》、第四名《ENDER LILIES》) 都已經很厲害了，那《GNOSIA》絕對值得一試。

Continue reading “日式人狼遊戲的回憶：《GNOSIA》玩後感及劇情探討”