顺便付一下大唐另一个对手-吐蕃帝国的资料
吐蕃（藏文：བོན་ཆེན་པོ།，威利：bon chen po，意为“大蕃”[1]）是历史上一个位于青藏高原的政权，由松赞干布到朗达玛延续两百多年。虽然有资料说明西藏早期已经建立了一个政治实体，但关于六世纪末以前的情况还没有确凿的历史证据。

637年，松赞干布率兵攻吐谷浑，并第一次与当时统治中国大部的唐朝发生冲突。640年，松赞干布派噶东赞迎娶唐朝的文成公主，文成公主携来了释迦牟尼佛像，并修建了饶摩齐（即今拉萨的“小昭寺”）。此外，松赞干布的另外三个妃子象雄公主、门妃迟江、惹拥妃等也分别创建了寺院，据说当时在整个西藏总共有108座佛庙。大约在641年的时候，松赞干布派噶东赞（即唐史中的禄东赞）到尼泊尔迎娶赤尊公主[3]，墀尊公主携来了西藏的第一尊佛像不动金刚佛，并修建了饶萨孜拉康（即现在的拉萨“大昭寺”）。[4] 大致上也在这个时候，从印度传入的密教和从汉地传入的大乘佛教，与西藏原有的“苯教”结合，而形成了藏传佛教。该教宣传宇宙大劫、一切无常、因果报应、生死轮回、修行解脱等教义，特别崇奉密宗，以无上瑜伽为最高修行次第，称为藏密。这是佛教传入西藏的第一波活动，宗教史学者将其称为藏传佛教的“前弘期”。松赞干布过世以后，吐蕃王朝与唐朝的关系时而紧张，时而修好。714年，吐蕃与唐朝议和，双方签订条约，规定以河源为界。未几战火又起，所以于730年又再度签约定赤岭为界。据《敦煌文选》和“达扎乐宫石碑”的记载：唐朝应允向吐蕃纳绢5万匹，以为岁绢。至唐代宗继位以后，停止向吐蕃纳绢。当时在位的唐金城公主之子，吐蕃赞普赤松德赞[5] 极为恼怒，遣部下达扎乐宫率20万之众于762年（安史之乱尚未完全平息）攻陷唐都长安，并于签约后退兵。赤松德赞（755年-797年）在位时吐蕃的国力达到鼎盛，除了大力发展佛教以外，还广拓疆土，疆界西至勃律、大食，东到川陇、南诏。

经过近两个世纪的战争，吐蕃和唐朝均感到疲惫。821年及次年，吐蕃和唐朝双方分别在长安、逻娑（即拉萨） [6] 会盟，达成“唐蕃甥舅和盟”。由于这次会盟是发生在唐长庆元年和二年，中国的史书称之为“长庆会盟”。记载这次会盟内容的“唐蕃会盟碑”共有三块，其中一块立于拉萨“大昭寺”前面[4]。在这次会盟所签定的条约中，吐蕃和唐朝两国的边界被确定，双方平等相待，互相承诺尊重对方的领土主权：
今蕃汉两国，所守见管本界，界以东悉为大唐国境，以西尽是大蕃境土，彼此不为寇、不举兵、不相侵。吐蕃王朝时期的西藏一直处于本土的苯教与新传入的佛教两股势力的争斗之中。吐蕃末期，苯教势力又占了上风。会昌元年（841年）可黎可足被反对佛教的大臣暗杀，其兄朗达玛被扶上赞普之位。朗达玛代表着反对佛教传播的势力。841年末代赞普朗达玛下令禁止佛教[7]，杀害僧人首领，强迫僧人还俗，关闭寺院，毁坏佛像、佛经，史称朗达玛灭佛。会昌六年（846年）达磨被僧人拉隆·贝吉云丹射死。达磨的两个儿子在父亲死后又为争夺赞普之位爆发内战，掌握军权的将领也随即发生混战，长达二十多年。统治集团的混乱又在咸通十年（869年）引起奴隶、属民的大起义，乾符四年（877年）起义军发掘赞普陵墓，逐杀王室和贵族，吐蕃王朝彻底崩溃，西藏二百多年的统一局面到此告终，西藏进入分裂割据的时代。[4] 如果从宗教发展的角度来看，这也是藏传佛教“前弘期”的结束。

这是一张公元820年的世界地图

与古羌人的关系
在青藏高原东南部地区，藏族早期的神话传说认为藏人起源于东方，即今四川、西藏、青海相邻地区。西藏的历史记载中还提到古羌人与西藏米博懂氏的血缘关系。故有观点认为藏族起源于羌族，即“汉藏同源”。原始藏族在汉文史籍中被称为发羌。在6世纪以前，整个西藏高原都分布著很小国家和部族，多说藏语，也有羌人与鲜卑人的后代。这些部落后来合并为12个小邦，其中以位于山南地区雅隆河谷的羌族的一支最为强大，不但统一了诸邦，而且后来还建立了吐蕃王朝。

据历史记载，从神龙元年（705年）至彝泰八年（822年），唐朝和吐蕃共会盟八次，其中第八次会盟碑至今仍矗立在拉萨大昭寺前。因为第八次会盟是在唐穆宗长庆元年至二年（821年至822年）进行的，所以也称为“长庆会盟”。盟文如下： 大唐文武孝德皇帝与大蕃圣神赞普，舅甥二主，商议社稷如一，结立大和盟约，永无沦替，神人俱以证知，世世代代，使其称赞。是以盟文节目，题之于碑也。文武孝德皇帝与赞（普）陛下，二圣舅甥，睿哲鸿被，晓今永之屯，享矜闵之情，恩覆其无内外，商议叶同，务今万姓安泰，所思如一，成久远大善，再续旧亲之情，重申邻好之义，为此大和矣。今蕃汉二国，各守见管本界，（洮泯以东，大唐封疆，其塞）以西，尽是大蕃境土，彼此不为寇敌，不举兵革，不相侵谋封境。或有猜阻，捉生问事讫，给以衣粮放归。今社稷叶同如一，为此大和。然舅甥相好之义，善谊每须通传，彼此驿骑，一（任常相往来，两路）蕃汉，并于将军谷交马，其绥戎栅已东，大唐只应，清水县已西，大蕃供应，须合舅甥亲近之礼，使其两界烟尘不扬，罔闻寇盗之名，复无惊恐之患。封人撤备，乡土俱安，如斯乐业之（恩，垂于万代，赞美之声），遍于日月所照矣。蕃于蕃国受安，汉亦汉国受乐，兹乃合其大业耳，依此盟誓，永久不得移易，（于）三宝及诸贤圣，日月星辰，请为知证。如此盟约，各自契陈。刑牲为盟，设此大约，倘不依此誓，蕃汉（背约破盟者，受其殃）祸也。仍须仇（报）及为阴谋者，不在破盟之限。蕃汉君臣，并稽首立誓，周细为文，二君之验，证以官印，登坛之臣，亲署姓名。手执如斯誓文，藏于玉府焉。

与大宋的关系建隆元年（960年），
北宋建立。由于大宋的所处的地区的缘故，与远离汉族地区的卫藏等地藏族各部联系交往较少，但与邻近汉族地区的甘、青、川、滇藏族各部则相对密切。大宋开国不久，原吐蕃帝国的一个部落首领的后裔唃厮啰在以邈川（今乐都）、青唐（今西宁）为中心的青海省湟水流域建立了政权。这是当时以藏族为主体的一个最大的地方封建政权。唃厮啰多次遣使向宋朝皇帝纳贡，并乞官职。明道元年（西元1032年），宋仁宗封唃厮啰为“宁远大将军、爱州团练使”，并给以优厚的俸禄。康定二年（西元1041年），大宋皇又封唃厮啰为“检校太保充保顺、河西等军节度使”。此后，唃厮啰的子孙继承人董毡、阿里骨、瞎征、陇等，世代均由大宋封官任职。唃厮啰部统治地区，隶属于北宋全国二十六路中的秦凤路。到政和六年（西元1116年），唃厮啰部地区全部改为宋朝的郡县。此外，宋真宗皇帝还于咸平四年（西元1001年）加封凉州（武威）藏族六谷部首领潘罗支为“盐州防御使兼灵州西面都巡检使”的官职。从西元11世纪开始，大宋在今甘肃省南部、河西走廊、青海省东部、西康等地区，大力推行屯田戍边的政策，以对付西夏的入犯，增加防卫力量。在此基础上，大宋向河西走廊等地的藏族部落拨发弓箭及其它武器，并招募藏族弓箭手，在藏人中建立类似大宋民兵的军事体制，以共同防御西夏人的入犯及袭扰。宋将王韶在今甘肃临夏、临洮一带大量开拓土地，招纳30多万藏人从事垦种。著名的茶马互市也在今四川雅安、甘肃临夏以及陕西一些地区的大宋专设市场上进行，开始了尔后数百年藏区马匹与汉区茶叶的经常固定交易。

是吗？但我看很多英文资料上也说的是3万左右

Eventually, though, a crisis developed when a pro-T*betan ruler came to power in the kingdom of Gilgit, located in the neighborhood of the modern city of the same name, now in Pakistan. The Chinese, after a number of unsuccessful attempts to rectify this unacceptable situation, finally sent a large army westward in 747 under the command of the famed Korean general Kao Hsien-chih who, in a dramatic campaign, remembered long after in both China and Korea, crossed the Pamirs and swooped down on an unsuspecting Gilgit. Demolishing the bridge across which T*betan reinforcements would have had to arrive, Kao beheaded selected pro-T*betan officials, and took the king and his T*betan wife prisoner -thereby ending Gilgit's flirtation with China's enemy, but failing nonetheless to head off the impending collision with the Arabs at Talas.
On the Arab side, our key informants on the Battle of Talas are the renowned historian Ibn al-Athir (1160-1233) and the comparatively unheralded al-Dhahabi (1274-1348). (Curiously, the most outstanding early Muslim historian, al-Tabari [839-923], has nothing to say about the Battle of Talas - indicating, perhaps, that from the medieval Islamic perspective, the battle seemed a rather peripheral encounter.) And they, along with Chinese sources, suggest that the clash of empires at Talas had its roots in a purely local quarrel, in the year 750, between the rulers of two petty kingdoms - Ferghana and Chach - that caused Ferghana to seek the military assistance of the Chinese. Kao Hsien-chih, now governor of Kucha, responded by besieging Chach, promising its king safe passage and then treacherously decapitating him. The son of the executed ruler, however, escaped and got word to Abu Muslim, the Abbasid governor of Khorasan. Sensing a golden opportunity to diminish China's political role in Central Asia, Abu Muslim quickly mustered his army at Merv — in today's Soviet Turkmenistan - added reinforcements from Tukharistan, a province in the north of today's Afghanistan, and crossed the Oxus to march to Samarkand. There, he rejoined the army of Transoxania under Ziyad ibn Salih, formerly the Umay-yad governor of Kufa in Iraq, and Ziyad took command.
The Chinese had mobilized - in concert with the troops of Ferghana - 30,000 men, according to Chinese accounts, 100,000 according to the Arabs and in July, 751, met the armies of Islam near the town of Talas or Taraz on the Talas River. A modern city of Talas can be found in the Kirghiz S. S. R., but medieval Talas probably lay nearer to present-day Dzhambul.
Chinese annals say the fighting lasted for five days, while Arabic records are inconclusive as to the duration. The end result of this epic encounter, however, is unanimously attested by our sources. The Arabs, aided by the Qarluq Turks, utterly destroyed the Chinese army. In the words of al-Dhahabi, "God cast terror into the hearts of the Chinese. Victory descended, and the unbelievers were put to flight"
o al-Dhahabi, the battle was won by the strategic acumen of Ziyad ibn Salih whose name will forever be remembered in connection with this thunderous Arab triumph beyond the Jaxartes. But the Chinese pin the blame on the Qarluq Turks, who, one account says, were "revolting" or "rebelling" against Kao Hsien-chih. In other words the Qarluqs deserted the Chinese coalition and changed sides in the midst of the action. In fact, the Qarluqs, far from mutinying during the engagement, were allied with the Arabs from the beginning and probably attacked the Chinese from the rear as part of a carefully prearranged battle plan - formulated, we may suppose, by Ziyad ibn Salih


Medievalists, military historians, Arabists, and Sinologists have debated the long-term political ramifications of the Battle of Talas. Never again, it is true, were the Chinese to play a significant role in Central Asia west of the Tarim Basin. Furthermore, this area, in which Buddhist, Zoroastrian, Manichaean, and Nestorian Christian influences had been strong, eventually became almost entirely Muslim. Thus we sometimes read statements to the effect that "those historic days determined the fate of Central Asia," to quote one modern authority.
Nevertheless, it would be an exaggeration to regard the overwhelming Arab victory at Talas as the sole cause of China's withdrawal from the western half of Central Asia, since, about the same time that Kao Hsienchih's troops were vanquished by those of Ziyad ibn Salih, China was experiencing trouble elsewhere on her borders - with the bothersome T*betans, the Uygur Turks in Mongolia, and in Manchuria the Khitan people, later to conquer much of North China. Khitan's troops, for example, thrashed a Chinese army near Ping-lu in 751 and in 754 Chinese forces suffered terrible defeats at the hands of the young kingdom of Nan-chao - today's Yunnan province - whose rulers were of Thai origin. The first of these two setbacks was sustained less than two months before the July debacle at Talas.
Simultaneously, China was being rocked by dissension within. Simmering opposition to the policies of Emperor Hsuan-tsung culminated in the revolt of 755, and the combination of external pressures and internal convulsions left China enervated and in a woeful state of decline. The defeat at Talas, therefore, did not by itself drive China permanently out of West Turkestan; it was a weakened China's manifest inability to bounce back after Talas that proved decisive in the end. In addition, to say that the battle was directly responsible for Central Asia's "turning Muslim" is to ignore the very real spiritual dimension in the region's gradual conversion to Islam -to wit, the strong inherent appeal of the Islamic faith, profoundly demonstrated over 13 centuries of world history.

那帖子里面提到的旧唐书记载的2万又是这么回事？

100000 according to the arabs, 这里说的是 both might be greatly exaggerated。

那是因为他们不懂中文 所以不能对中国的史料进行进一步的研究
但作为中国人， 不可能不把官史旧唐书里面的数据考虑进去吧

城傍兵是什么？求解